Gestion Del Agua

G. M.
Tarekul Islam
Shampa
Ahmed Ishtiaque Amin Chowdhury Editors
Water Management:
A View
from Multidisciplinary
Perspectives
8th International Conference on Water
and Flood Management
Water Management: A View from Multidisciplinary
Perspectives
G. M. Tarekul Islam · Shampa ·
Ahmed Ishtiaque Amin Chowdhury
Editors
Water Management: A View

from Multidisciplinary
Perspectives
8th International Conference on Water
and Flood Management
Editors
G. M. Tarekul Islam Shampa
Institute of Water and Flood Management Institute of Water and Flood Management
(IWFM) (IWFM)
Bangladesh University of Engineering Bangladesh University of Engineering
and Technology (BUET) and Technology (BUET)
Dhaka, Bangladesh Dhaka, Bangladesh

Institute of Water and Flood Management
(IWFM)
Bangladesh University of Engineering
and Technology (BUET)
Dhaka, Bangladesh
ISBN 978-3-030-95721-6 ISBN 978-3-030-95722-3 (eBook)

https://doi.org/10.1007/978-3-030-95722-3
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Switzerland AG 2022
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
This book Water Management: A View from Multidisciplinary Perspectives presents

the collection of selected papers from the 8th event of the International Conference
on Water and Flood Management (ICWFM) held during March 29–31, 2021. This
year, the conference added a new dimension not only because the program was held
online due to the COVID-19 situation but because the pandemic taught us how to act
when natural disasters and pandemic-like situations occur concurrently.
Said that, organizing the conference in virtual platform was less difficult compared
to the challenges of selecting top articles from more than 200 scientific works
presented in the conference. Thanks to the members of the Scientific Committee and
Reviewers whose relentless effort has made this possible. It is gratifying to note that
our initiative has brought together a multidisciplinary and global team of national and
international academicians, researchers, experts, as well as practitioners contributing
chapters to the book on the diverse physical, environmental, socio-economic, and
institutional issues concerning Water Management. Our objective, and indeed, the
driving force behind the book, has been to bring the current water management issues
and knowledge throughout the world into the forefront of the scientific community
and policymakers. We hope all this endeavor will prove valuable to academicians,
policy planners, and development practitioners alike.
Finally, thanks to the Springer team for publishing this volume for two succes-
sive years. This particularly encourages early career researchers to contribute to the
conference.
Dhaka, Bangladesh Prof. G. M. Tarekul Islam

Shampa
v
Contents
Floods and Drainage

Causes and Management of Damaging Flood Incidences in Rapidly
Urbanizing Areas of Kathmandu Valley: A Case Study of Flood
Event in Bhaktapur District, Nepal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Purnima Acharya and Ashutosh Shukla
Urban Drainage Study for Gopalganj Pourashava Considering
Future Climate Change Impacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Faruque Abdullah, A. K. M. Saiful Islam, Afsara Tasnia,
G. M. Tarekul Islam, Sujit Kumar Bala, and Nahruma Mehzabeen Pieu
Flood Propagation Processes in the Jamuna River Floodplain
in Sirajganj . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Ashik Iqbal, M. Shahjahan Mondal, M. Shah Alam Khan,
Hans Hakvoort, and William Veerbeek
Co-creation of Flood Mitigation Technologies in Bangladesh
to Strengthen Community Resilience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Nadia Nowshin, M. Shah Alam Khan, Hans Hakvoort,
William Veerbeek, and Chris Zevenbergen
Urban Waterlogging Risk Profiling: The Case of Khatunganj
Wholesale Commodity Market, Chattogram . . . . . . . . . . . . . . . . . . . . . . . . . 93
Tasnim Alam Nishat, Dewan Salman Sunny, Rifat Talha Khan,
Md.Reaz Akter Mullick, and Piyal Datta
Hydrometeorological Hazards and Risk

A Remote Sensing-Based Approach for Analysis of Dry and Wet
Periods of Bangladesh Based on Standardized Precipitation Index
During 1981–2020 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Saumik Mallik
vii
viii Contents
Indigenous Knowledge and Practices of the Small Ethnic

Communities of Asia–Pacific Island Countries in Facing
Hydro-Meteorological Hazards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
Mahfuzul Haque
Driving Factors of Destination Choices Due to Riverbank Erosion
Along the Brahmaputra River . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Sahika Ahmed and Sonia Binte Murshed
Bivariate Drought Risk Estimation Using a Multivariate
Standardized Drought Index in Marathwada Region, India . . . . . . . . . . . 173
Rajarshi Datta and Manne Janga Reddy
High-Quality Historical Flood Data Reconstruction in Bangladesh
Using Hidden Markov Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Max Mauerman, Elizabeth Tellman, Upmanu Lall,
Marco Tedesco, Paolo Colosio, Mitchell Thomas, Daniel Osgood,
and Arifuzzaman Bhuyan
Rivers, Coasts and Estuaries

Impact of Coriolis Force on the Flow Field and Sedimentation
in Ideally Shaped Tidal Basins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
Nazeat Ameen Iqra, Mohammad Asad Hussain, and M. Shah Alam Khan
The Impact of Small Tributaries Flood in the Braided Plain
of Large River . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Md. Manjurul Hussain, Shampa, Juwel Islam, Md. Shibbir Ahmed,
Md. Ashiqur Rahman, and Md. Munsur Rahman
Water Infrastructure and Development

Hybrid Coast Protection Approach in Bangladesh: A Case Study
on Effectiveness of Small-Scale Forest in Reducing Surge Induced
Inundation and Supporting Local Livelihoods . . . . . . . . . . . . . . . . . . . . . . . . 251
Mita Kazi Samsunnahar
Assessing the Consequences of Large-Scale Stabilization
of the Padma River on Its Flow Hydraulics Using a Combined
1D-2D Hydrodynamic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
Subir Biswas and M. Shahjahan Mondal
Water and Livelihood Security

A Sustainability Index for Assessing Village Tank Cascade Systems
(VTCs) in Sri Lanka . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
E. M. G. P. Hemachandra, N. D. K. Dayawansa,
and Ranjith Premalal De Silva
Contents ix
An Agent Based Model of Mangrove Social-Ecological System

for Livelihood Security Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
Shamima Airin Sweety, M. Shah Alam Khan, Anisul Haque,
and Mashfiqus Salehin
Drought Management by Integrated Approaches in T. Aman Rice
Season to Escalate Rice Productivity in Drought Prone Regions
of Bangladesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
Debjit Roy, Md. Belal Hossain, Mohammad Rezoan Bin Hafiz Pranto,
and Md. Towfiqul Islam
Actual Evapotranspiration Estimation Using Remote Sensing:
Comparison of Sebal and Metric Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
Sumit Kumar Saha, Rubel Ahmmed, and Nasreen Jahan
Floods and Drainage
Causes and Management of Damaging
Flood Incidences in Rapidly Urbanizing
Areas of Kathmandu Valley: A Case
Study of Flood Event in Bhaktapur
District, Nepal
Purnima Acharya and Ashutosh Shukla
Abstract The ever-growing incidences of urban flooding that were previously

anomalous have now started affecting many parts of Kathmandu Valley on a yearly
basis. This study aims to understand the causes of urban floods and the adap-
tation/mitigation interventions being implemented through a case study of recur-
rent urban flooding in Bhaktapur District, Nepal. A Land Use Land Cover Change
(LULCC) map to trace the local LULCC of the flood-affected area and precipitation
data of Bhaktapur city was analysed to examine its contribution to the flood. House-
hold surveys, FGD and key informant interviews were conducted to understand the
mitigation and adaptation practices at household, community and national levels.
The research revealed that the flooding impact has been increasingly severe in recent
years, mostly due to poor mitigation efforts. Poor storm water management aggra-
vated by disorganized human settlements in the right of way of the river contributes
to scale up the infrequent overflows to overt flooding. Therefore, it is important to
expand our understanding regarding causes and consequences of flooding events to
inform urban planning. Considering current and potential adaptation strategies, this
research makes suggestions for effective adaptive measures and adds to a global
conversation about the future of sustainable cities.
Keywords Urban Flooding · LULCC · Climate Change · Storm water · Drainage
1 Introduction
The annual flood brought by heavy monsoon precipitation results in enormous

damage to lives and properties, making regions of southern Asia vulnerable to floods
(Mirza 2011). In addition, population growth and changes in land use have also
increased human vulnerability to floods (Dewan 2015). The changes in land use not
only bring changes in the geomorphology but also change the infiltration capacity
of soils, which leads to increase in the amount of overland water flow causing floods
P. Acharya (B) · A. Shukla

Nepal Engineering College, Bhaktapur, Nepal
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 3

G. M. Tarekul Islam et al. (eds.), Water Management: A View
from Multidisciplinary Perspectives,
https://doi.org/10.1007/978-3-030-95722-3_1
4 P. Acharya and A. Shukla
downstream. In addition, the changes in natural drainage and increased burden on

existing drainage system, as a result of population growth, increase the likelihood
of the system being overwhelmed (Salike and Pokharel 2017). With extreme rainfall
events having more recurrent trends in the recent years (IPCC 2014), incidences of
both pluvial and riverine flooding have been exacerbated.
Flooding has become a major problem in Kathmandu Valley, Nepal in recent years.
Rapid urbanization and subsequent human led changes in the urban landscape has
increased the imperviousness exacerbating the flood frequency. The flood incidences
over the last one decade have increased in areas with no flooding events in the past, that
also includes areas with relatively low precipitation in the catchment. In Hanumante
River of Bhaktapur district, there have been two flood events within the four years
producing widespread damages—the first one occurring on August 27, 2015 and the
second on July 12, 2018 (Prajapati et al. 2018). This study mainly focuses on the
flood incidence of July 12, 2018.
While incidences of urban flooding are on the rise all over Nepal, it remains one of
the least explored areas in the field of disaster management (NWCF 2009). The chal-
lenges to flood hazard mitigation are essentially about instituting good governance
(Rijal et al. 2018). The policy, as it exists, does not allow the government to impose
necessary restrictions on private property, thus proving ineffective in preventing the
unsustainable expansion of urban sprawl.
This research analyses the factors and processes accumulating in the urban areas
responsible for the genesis of the flood event of July 12, 2018 in Hanumante River
of Bhaktapur, Nepal. It also analyses the strings of adaptation and mitigation efforts
implemented by the community and different levels of government. It also aims to
draw key gaps in the policies, regulations and practices leaving the urban flooding
unaddressed in the rapidly changing urbanscape and urban development plans.
2 Methodology
2.1 Study Area
The study area (Fig. 1) of this research lies in Madhyapur Thimi Municipality (MTM)
in Bhaktapur District. Geographically, MTM lies at 27˚40 00 to 27˚42 00 N lati-
tudes and 81˚22 30 to 85˚25 00 E longitudes, at an average elevation of 1320 m from
the mean sea level. Much of the landscape of the municipality is plateau-like, located
between Manohara and Hanumante Rivers, which includes conspicuous upland and
lowland areas.
Hanumante River flows from north-east to south-west of Bhaktapur District with
an average width ranging from minimum of 10 m in the dense urban sprawls to
maximum of 20 m in peri-urban and rural areas. The catchment of the river is 143 km2
stretching across 23.5 km (Sada 2012). The river includes numerous tributaries and
among them Ghatte Khola, Kasan Khola, Kalighat Khola and Kaalcha khola are
Causes and Management of Damaging Flood … 5
Fig. 1 Study area
important tributaries which join the river at different stretches. Specifically, the low
land areas were in the focus of the study because these were the areas inundated in
the flood event of July 12, 2018.
This research has used a case study approach and multi-perspective analysis:
geographic, geomorphic, meteorological and socio-economic in unpacking the flood
incidence of July 12, 2018 in Hanumante River that inundated Madhyapur Thimi
Municipality. The following methodological tools were used.
2.2 Reconstruction and Mapping of the Studied Flood Event
The flood event of July 12, 2018 was reconstructed using secondary data collected
from different published/unpublished sources and primary information was collected
through the questionnaire survey with thirty-five people who faced the event. The
primary information was also received from those associated to different social and
humanitarian organizations and local governments who were involved in rolling
rescue/relief operations following the event and mitigation efforts in the aftermath of
the event. The information collected was validated with the recorded information of
local government, civil society organizations, red-cross and district administration.
2.3 Reconstruction of Historical Events
Hanumante watershed is ungauged therefore longer-term hydrological data and

records of flood history was not available. The Disaster Risk Reduction portal of
the Ministry of Home Affairs which is responsible for maintaining records of the
disaster event across the country did not have any record of flood events in Madhyapur
Thimi and/or Hanumante. Similarly, Municipal Office, Department of Hydrology and
Meteorology (DHM), Department of Irrigation and Water Resources (DIWR), Water
and Energy Commission Secretariat (WECS), District Administration Office, Nepal
Red Cross Society, Bhaktapur and District Police office had no archived information
regarding historical flood events. In the absence of hydrological stations at any of its
reach in the Hanumante River and unavailability of water flood records, hydrological
analysis could not be done, so, the records of flood history were developed using
information from the memory of the people.
2.4 Meteorological Data Analysis
The study used rainfall data maintained by the DHM to analyse the rainfall trend
in the areas. For this purpose, daily rainfall data of the time period 1971–2018
of the nearest meteorological station (Nagarkot index no. 1043, Bhaktapur index
no. 1052, and Changunarayan index no. 1059) were analysed using RCLIMDEX
software (Zhanag and Yang 2004). These stations were selected by drawing Thiessen
Polygon to establish possible contributions in Hanumante catchment. Out of three
rainfall stations closer to Hanumante basin, the data of Bhaktapur station were only
found to be homogeneous while conducting RH test and hence was only used for
rainfall analysis.
RClimDex is a software package which is useful in calculating the climate
extremes indices, required to monitor and detect climate change (Zhanag and Yang
2004). RClimDexuses R platform was used to perform and check the data quality
before computing the indices (Zhanag and Yang 2004). This study uses selected
precipitation indices for the rainfall analysis and conducts quality control which
checks the data input errors which also includes missing data as well. The RHtest
software was applied for homogeneity testing of the rainfall data as it does not require
any reference series and is available freely (Wang and Feng 2013).
2.5 Map Analysis
The study has analysed the changes in land use and land cover around Hanumante
River in Madhyapur Thimi Municipality (MTM) where time series analysis was done
from the late seventies to look into the decadal change. Due to limitation of the high-
resolution images for detailed LULCC analysis in the small area along Hanumante
River, this study uses aerial photographs of 1979 and 1992 which were available with
the Department of Survey, Government of Nepal and the Google earth images of 2005
and 2018 which were downloaded and refined for the year 2005 and 2018. These
images were used to develop the LULCC map for the lowlands of Madhyapur Thimi
and Suryabinayak Municipality, that lies within 500 m width from the Hanumante
River for the stretch of 7.13 km covering 7.34 km2 , to know the increase in built-up
area along the Hanumante River, which is prone to flooding.
After georeferencing the aerial photographs in ArcGIS, digitization of the photo-
graph was done using four classifications to look into the changes in the land use
and land cover: (i) vegetation, (ii) water body, (iii) settlements and (iv) farmland and
open space. Similarly, digitization of the Google earth image was done using the
same classification and decadal land use and land cover changes were assessed.
2.6 Focus Group Discussion
Experiences and perceptions of the flood event, the changing rainfall pattern, causes
and impacts of the flood events were collected from the urban poor, slum and squatter
settlers, old inhabitants and recent migrants with their dwellings close to the river
course through FGDs, with the following two groups having ten people in each of
the groups. Series of questions were asked and documented.
i. Old habitants living in the upland of MTM but having farms in low land and
ii. Old and new migrants living in the lowland of MTM who at least experienced
the flood incidence of July 12, 2018.
2.7 Key Informants’ Interviews (KII)
Fourteen KIIs were conducted to collect information pertinent to plans, policies

and strategies of the government; social, economic, infrastructural and environ-
mental changes in the area and planned and autonomous adaptation practices rolled
at different levels. The key informants included personnel working in KVDA,
DHM, DOI (WRRDC), Smart Panni, HPCIDBC, MTM, Bhaktapur Municipality
and Red Cross Society, including independent professionals and academicians and
researchers.
2.8 Physical Observation of the Study Area
The land use pattern and the corridor development work in progress implemented
by the High-Power Commission on Integrated Development of Bagmati Civilization
(HPCIDBC) were observed as part of the study. At some sections the original width
of the river was narrowed as a result of development of the road corridor, which
was indicative of the encroachment in the river even by the government supported
development programmes.
3 Results
3.1 Historical Flood Events in Hanumante River
The focus group discussion, household interviews and key informants’ surveys
showed that the lowland areas of MTM have had a past history of getting flooded
intermittently at an interval of a few years but the magnitude of damage produced
by the recent floods including on July 12, 2018 was unprecedented. In the past, there
were no disturbances on the passage of flood and therefore the water level in the river
receded within a few hours to few days’ time. But the floodwater on July 12, 2018
stood in the area for more than 24 hours showing significant change in the recession
of flood flow.
3.2 Analysis of Climatic Causes of the Flood
In order to analyse if the variability in rainfall has been responsible for the flood
event, rainfall records of Bhaktapur meteorological stations were analysed using
RCLIMDEX software. The catchment of the Hanumante River is 143 km2 and almost
all of the surface water of Bhaktapur is drained through this river. Therefore, when
it rains covering the entire catchment the volume of the water in the river rises,
producing potential risk of flooding and inundation in the downstream areas.
3.2.1 Rainfall Frequency Analysis
Daily rainfall records of Bhaktapur Station for the period 1971–2018 were subjected
to frequency analysis using Gumbel’s Type I Extreme Distribution Function to
estimate the return period of rainfall of different magnitude (Table 1).
The stations at Bhaktapur, Changunarayan and Nagarkot recorded 24-h rainfall
of 129.6, 30, and 117 mm, respectively, on July 12, 2018. As the Bhaktapur station is
considered representative to the study area and therefore contributor to the genesis of
the flood event on the day, the return period of the event is calculated to be 8.4 years.
This shows that the likelihood of recurrence of the flood event of the magnitude of
July 12, 2018 is once in 8.4 years (Fig. 2).
Table 1 Estimated 24-hours

S. No. Return Period (years) Rainfall (mm)
rainfall for different return
periods 1 2 73.3
2 5 110.99
3 10 135.95
4 50 190.87
5 100 214.08
Fig. 2 Likely 24 hours rainfall at different return periods in Bhaktapur

3.2.2 Analysis of Rainfall Trend
This study maintained that one of the causes for the flooding event was extreme
rainfall event. The five indices used in the study provide opportunity to look into
antecedent rainfall conditions and persistence in the event, both proving bases to
logically unpack the extremity of the event.
i. Trend of Annual Total Rainfall
The annual total rainfall analysis revealed a decreasing trend, but the trend
was found non-significant as the p-value is 0.075 (i.e. non-significant at 5%
level of significance). The decreasing trend in annual total rainfall would mean
lower possibility flooding occurrences in the study area in future. However,
relying on this observation to make any assessment of flooding trends in the
study area would be grossly unrealistic for the fact that flood events are often
results of extreme events and the runoff producing potential of the catchment
and not necessarily the contribution of total annual rainfall in the catchment.
ii. Trend of Extreme Values
In order to analyse the likely contribution of extreme values responsible
for the genesis of the flood event of the day, the highest daily rainfall amount
(RX1day) and highest 5-days consecutive rainfall amount (RX5day) of Bhak-
tapur Station were subjected to trend analysis. The analysis revealed an
increasing trend (p = 0.521) of highest daily rainfall while decreasing trend
(p = 0.719) in 5-days consecutive rainfall total. However, this trend cannot be
relied upon as both of the observed trends are very weak and non-significant.
iii. Trend of Consecutive Wet Days (CWD)
The observed trend in maximum consecutive wet days was noted to be
decreasing significantly (p = 0.041) in the study area. This trend signals
lowering of persistence in rainfall in future times, even in the monsoon. Lower
CWD in the area even if there is not much change in the total amount of
rainfall would mean lower runoff concentration to produce peak flow because
runoff yield from catchment depends on antecedent moisture condition in the
catchment besides the amount of the rainfall. Similar observations have been
observed of the people living in the study area for a longer time, who revealed
that monsoon rainfall in the area is no longer persistent. Number of days without
rainfall in between two successive rainfall events has increased, unlike in the
past.
iv. Simple daily intensity index (SDII)
In order to look into the trend of SDII in the study area, the SDII values of
each year over the past 47 years (1971–2018) were plotted and it revealed a
decreasing trend in the study area, which was non-significant (p = 0.087). This
trend of SDII in the study area revealed decreasing rainfall amounts on the wet
days to produce concentrated runoff causing flooding in the study area.
All the rainfall trend plots are shown in Fig. 3.
PRCPTOT: Trend of annual total precipitaƟon
RX1day: Trend of highest 1-day rainfall StaƟon
Rx5day: Trend of highest 5-day rainfall StaƟon
CWD: Trend of consecuƟve wet days
SDII: Simple daily intensity index

Linear least square
Locally weighted Linear
regression
Annual data
Fig. 3 Trend of rainfall analysis
3.2.3 People’s Perception on Climatic Trend
People in the area recall significant shifts in the pattern of rainfall and seasonality
indicating that the past moderate and persistent monsoon rainfall had mostly been
of short duration with high intensity. Increased occurrence of high intensity rainfall
event in the monsoon was attributed by most people to the cause of flooding in Hanu-
mante River and inundation in the surrounding areas, however out of 35 local people
who were interviewed, only a small number of people (less than 20%) pointed the
rainfall to be the only cause of the flooding. They rather attributed changes in the land
use, particularly changes brought in the natural drainage pattern as a result of rapid
infrastructure development to be the additional cause. This attribution signifies that

they do distinguish rainfall amount and the surface water flow that the given amount
of rainfall produces. This perception of rainfall excess clearly signifies people’s
knowledge of the changes occurring in the landscape and hydrological processes as
a result of population growth and densification of human built infrastructures.
3.3 Analysis of LULCC
The study attempted to relate LULCC as a possible causal factor combining with
rainfall for producing the flood event of July 12, 2018 in Hanumante River. The
LULC maps of the area for different time periods covering a stretch of 500 m width on
both sides of the river in Madhyapur Thimi Municipality were prepared. In addition,
LULCC map of Kathmandu Valley was also prepared to relate the urbanization trend
and related LULCC in the study area with the general trend in Kathmandu Valley.
3.3.1 LULCC of Kathmandu Valley
Land use map of Kathmandu Valley was prepared in Google Earth Engine (Gorelick
et al. 2017) for the year 1988, 1998, 2009 and 2018 to assess the general trend of
urbanization in the valley and to compare the resulting changes with the changes in
the study area (Fig. 4). This was needed because LULCC in the study area cannot
be looked at in isolation.
The LULCC analysis of the Kathmandu Valley shows that between 1988 and 2018,
the size of the -developed area increased by 625% (Table 2). Much of this growth
was found concentrated in three major urban centres: Kathmandu Metropolitan City,
Lalitpur Metropolitan City and Bhaktapur Municipality. Human settlements were
also found to be rapidly expanding in the peripheral areas, radiating to all directions,
from the urban core mostly after the late nineties with 165.60% growth in the devel-
oped area from 1998 to 2009. This clearly reveals significant growth in the developed
area in Kathmandu Valley over the last three decades.
3.3.2 LULCC of Study Area
In order to assess the land use change in the lowlands of Madhyapur Thimi that was
engulfed into recurrent flooding and inundation on July 12, 2019, 7.34 km2 of area
that extended 500 m on either side of the Hanumante River for a stretch of 7.13 km,
one side in Madhyapur Thimi and on the other side Suryabinayak Municipality,
was analysed (Fig. 5). The periods selected for analysis included 1979, 1992, 2005
and 2018, to cover a time span of four decades (Fig. 6) and maps were prepared in
ArcGIS. There has been significant expansion in the built-up area over the period 0f
2005–2018 (Table 3), it increased from 84.9 to 389.330 ha. On the other hand, there
Fig. 4 Land Use Land Cover Change map of Kathmandu valley

Table 2 Land Use Changes in Kathmandu Valley (1988–2018)

Area (ha)
Year/Categories Built-up area Forest Other (Agriculture, barren, open space)
1988 2023.74 61,379.73 29,336.76
1998 4212.81 43,420.77 45,106.65
2009 11,189.34 35,886.87 45,664.02
2018 14,678.71 39,475.98 38,586.84
Fig. 5 The site location for LULCC in MTM

Fig. 6 LULCC map around Hanumante River
has been reduction in the area under agricultural use, water bodies and forest. This
resonates with the information provided by the people that the extent of damages
faced by the people in the flood event of July 12, 2018 was essentially due to growth
in human settlement in close proximity of the river and not by the magnitude of the
flood itself.
Table 3 Land use changes in the study area over 1979–2018

Area (ha)
Year/Categories Settlement Forest River Road Cultivation/barren land
1979 19.00 23.73 41.33 6.24 643.34
1992 54.64 19.48 30.72 6.31 622.49
2005 84.96 14.69 15.32 6.31 612.36
2018 389.33 8.81 11.90 17.27 306.34
3.3.3 People’s Observation on Urbanization vis-à-vis Trend of Flooding
This section presents the observation of the people living in the study area regarding
the pattern of flooding and inundation in relation to urbanization and changes in the
landscape along Hanumante River.
i. Growth in the Population and Human Settlements
The high rate of migration, development of roadways and conversion of
agricultural land have developed physical barriers for surface water flowing
back into the river during heavy rainfall. This has resulted in inundation for
prolonged duration in the human settlements on both sides of the river. Nearly
67% of the respondents blamed the government for not having strict policies
to regulate the land near the rivers.
ii. Encroachment into the Flood Plain
The key informants identified significant encroachment on the right of way
of the river. The area adjoining the river was previously a farmland that had
always been the flood plain of Hanumante River, but, at present, the entire area
is covered by new settlements and commercial establishments. Unregulated
conversion of the farmlands into settlements led it to emerge close to the river’s
waterway. Also, the natural barriers of endemic vegetation were gradually
removed that resulted in floodplain and the riverbanks becoming weak, causing
significant erosion due to bank cutting and sloughing. The sediments entered
into the river and were deposited on the bed, causing the river bed to rise. The
problem was aggravated when riverbank was used for dumping solid wastes
by the municipal authorities of Bhaktapur and Madhyapur Thimi.
iii. Poor Storm water Management Practices
People attributed the problem of flooding and inundation to their inability
to evacuate the excess water from the rainstorms quickly. With the reduc-
tion in the open spaces and porous land surface, the existing drainage system
gets overwhelmed even with rainfall of moderate intensity. The overland flow
from paved areas and human settlements, which generated suddenly following
the rainfall event, when added to the river flow at lower reaches, aggravates
flooding.
iv. Infrastructures Development without Drainage Considerations
Construction of new roads and other human built infrastructures brought
changes in the natural drainage system and the routes for the flow of excess
water in the study area. There are several bridges along the Hanumante River
and at these bridge crossings, the river section has been narrowed which blocks
the natural flow of the river.
v. Erosion of Traditional Systems and Practices of Water Management
The respondents revealed that in the past, there were a number of ponds,
natural drains, canals and open areas that buffered the flood water reducing
damage. Drainage and excess water from the stone spouts went to the ponds
downstream and during rainfall the ponds acted as a storage that was used
in the pre-monsoon period for agriculture (Molden and McMahon 2019). In
addition to this, surface runoff was also channelled into these ponds. These
traditional systems and water management practices are now on the verge of
extinction and the pace of development of new infrastructures hardly respects
their existence and accrue importance.
vi. Landscape Level Changes Caused by Land Development
Local governments allow land developers to extract sand from hillocks on
the river terraces in the valley, which is also common in Madhyapur Thimi.
The damage caused to landscape and the environmental damage resulting from
such practice of land development, which is rampant at present, is difficult to
reverse. The sandy hillocks in the landscape have an important role in sustaining
local hydrology for their ability to store water during and after rainfall. Highly
permeable sand formation retains the water which would regulate runoff. The
loss of the sandy hillocks in the landscape has been one of the reasons for
urban flooding, particularly in the catchment of Hanumante River in Bhaktapur.
With the loss of more porous sub-surface sandy formation rainwater moves out
quickly to produce floods and inundation on the downstream.
3.4 Management of Flood
Management of flood involves minimizing the negative impacts, but they are difficult
to eliminate altogether (Khanal 2018). Therefore, resilient adaptive strategy and
mitigation efforts rolled at various levels is a cornerstone for minimizing loss and
damage. In addition, it is imperative to explore sustainable solutions with regard to
management techniques.
3.4.1 Adaptation and Mitigation Efforts Rolled at Household

and Community Level
As stated in the preceding sections, the study area had been experiencing incidences
of recurrent flooding in the past. Therefore, the case flood incidence was not new
but the intensity and duration of flooding and inundation caused by it and the losses
and damages suffered were alarming. The event has made people more cautious of
repetition of events of similar or larger magnitude in the future.
People revealed that the magnitude of flooding and inundation, at least those
that had occurred in the recent past including the event of July 12, 2018, has been
so large and sudden, that this leaves very little scope at their end to cope with or
adapt to them. People think this would require larger efforts and investment which
would be beyond their reach and they foresee the importance of collective efforts
and investment made by local governments. Nonetheless, some of the measures that
they have started rolling at their end following the event to develop their adaptive
capacity include:
• Shifting to other areas or at least shifting to live on the 1st and 2nd floor of the
house.
• Raising the plinth level of new houses to ensure that the floodwater would not
enter the house.
• Investing to develop raised masonry boundaries, to function as flood walls.
3.4.2 Adaptation and Mitigation Rolled at Local Government Level
Some of the initiatives that have started at the local level following the event have
been: (i) removal of silt and cleaning of the river bed and banks annually with the
aim of maintaining the river cross-section to allow the floodwater to pass the river
channel quickly, and (ii) strengthening riverbanks by developing masonry and vege-
tated revetment at critical sections and strengthening the ghats and (iii) development
of the road corridor, under the funding of the High-Power Commission on Integrated
Development of Bagmati Civilization (HPCIDBC) which involves the development
of a 20 m wide river followed by 20 m wide road on both sides, and construction of
flood wall and retaining wall to strengthen the bank.
The hydrological report of this project reveals that the hydraulic calculation for
flood is carried out for the 2- and 5-year return periods at the cross-section and bridge
locations along the reaches of all rivers. Though the actual impact of the project is yet
to be seen, the real risk is the infrastructure development under the project producing
large and irreversible changes in the river environment. When inquired, 83% of the
respondents opined that the corridor project development will address the problem
while 17% of them maintained that the river width will be further narrowed, thus
limiting the capacity of the river to dispose-off the runoff that often produces from
extreme rainfall events. One obvious consequence that can be thought of is the
disconnection between surface water and groundwater in the river environment as a
result of creation of a physical barrier on the riverbank.
3.4.3 Adaptation and Mitigation Rolled at the Central Government

Level
The past flood events and the one of July 12, 2018 does not seem to have produced
enough concern at the level of central government agencies, those entrusted with
the responsibility of decision making and issue policy directives, that urban flooding
and inundation is an impending hazard risk in the urban areas and the consequences
could be grave if it remains unaddressed. Their focus continues to be on designing and
rolling infrastructure solutions to the problem. One such example which is currently
underway in the area is urban drainage development involving construction of trunk
sewer lines on both sides of the river collecting sewage disposal of the household.
The project is part of the urban water system improvement project of Kathmandu
Upatyaka Khanepani Limited (KUKL), implemented through a Project Implementa-
tion Directorate (PID). The design of the project does not involve any focus on collec-
tion and management of storm water, which is the cause of flooding and inundation
in the area.
4 Discussion
The flood event of July 12, 2018 was the result of several factors and processes,
all combining to produce the flood event of the day. The rainfall trend analysis
demonstrated an increase in the high intensity short duration rainfall events which
also corroborated with the people’s perception. With the increase in urbanization, the
events of flooding are seen to be more deleterious and detrimental to the inhabitants
in the floodplain. The analysis also shows that the likelihood of recurrence of the
flood event of the magnitude of July 12, 2018 once in only 8.4 years. Given the fact
that Bhaktapur station has historically witnessed the highest daily rainfall amount
of 260 mm recorded in 1990, occurrence of rainfall events of this magnitude could
produce even worse floods in the future if not addressed timely. This observation
must draw the attention of local governments and the agencies responsible for the
development and management of urban infrastructures and services. The Weather
Forecast Division of Nepal (DHM) agrees that events of urban flooding, such as
the one that occurred on July 12, 2018, are not solely the result of weather-related
causes. Most of the time the genesis of urban flooding lies on poorly developed urban
infrastructures and services.
The scope of mitigation and adaptation at the household level is much limited as
the recurrent urban flood events require larger scale efforts, involving a multi-pronged
approach of storm water management, improvement of drainage systems and regu-
lation of infrastructure growth in the urban areas. Contrarily, the local government
in the existing arrangement bears responsibility to roll preparedness but does not
seem to have any organized and coordinated preparedness measures. Development
of river corridors along Hanumante River undertaken by HPCIDBC was identified
as the only major activity in the floodplain, although the primary objective of this
was to develop road corridors and not for flood control.
The issues of disaster risk reduction and management (DRRM) remained outside
the scope of urban development policies until the proclamation of Hyogo Framework
(2005–2015) and Sendai Framework of Action (2015–2030). But, the attention has
been only on the high intensity and episodic disasters, small yet recurrent disasters,
such as urban flooding and inundation, continue to remain outside the radar of DRRM
especially in context of Nepal. One of the hard facts projected by the study is the lack
of attention and priority for urban flooding and inundation in the existing regulatory
codes and practices of Nepal. Nepal’s Urban Development Policy (2007) and Urban
Development Strategy (2017) considers ‘resilience’ as the guiding principle of urban
development and emphasizes on sensible urban development, and these policy direc-
tives do set conceptual clarity for sustainable urban development. However, they do
not provide for an operational mechanism to translate the policies into action on the
ground. The land use zonation (LUZ) maps which are basic instruments to regulate
the land use in the urban and rural areas have not been developed at the level of most
municipalities. Even in Kathmandu Valley where these maps have been developed,
it is yet to be implemented effectively. This is mainly due to the fair amount of ambi-
guity in the existing regulatory laws to impose restrictions on the development and
uses of private land other than those for designated purposes.
4.1 Sustainable Solutions to Mitigate Urban Flooding
Kathmandu was once a city of culturally and religiously significant ponds that enabled
urban development. These traditional systems are rapidly decreasing as the new
development did not respect their existence, to add on, the new development did not
care for the development of additional drainage facilities to compensate for this loss.
This has certainly produced a regime of increased flooding and inundation in urban
areas. The local knowledge that informed these ancient structures can be adapted to a
modern context by improving drainage systems, groundwater recharge and rainwater
storage. The use of traditional water storage systems in new development plans can
have the potential of adapting local knowledge systems to solve this modern problem.
Also, the dissipation of rainwater, especially its use to recharge groundwater, should
be part of the mitigation plan for urban flooding problems.
There is no doubt that the change in local hydrology due to urbanization has
contributed to an increase in the frequency of damaging flood incidences in Hanu-
mante River and the damage suffered by the people. This change in the local
hydrology is also exacerbated by other changes in land use such as the switch from
rice to vegetable cultivation and the practice of increasingly leaving land fallow, as
land without a cover crop does not hold water, it acts like a pavement during the
deluges of the monsoon and disrupts the recharge cycle (Wrobley 2020). Therefore,
planting rice could also be one of the solutions to hold the water.
For enabling resilient urban living, urban flooding and inundation should both
be included in the urban development policy and programming. One of the major
clauses to be included is the river waterway demarcation in order to avoid further
encroachment along the river as well as maintenance of natural drainage in the urban
areas. Planning and regulatory capacity of agencies, including local government, to
regulate haphazard urbanization should be developed to restrict urban areas from
turning into disaster hotspots.
5 Conclusion
The study intended to look into the factors and processes that were responsible for
the onset of the flooding event of July 12, 2018. This involved a multi-perspective
analysis involving analysis of rainfall data and LULCC in the catchment and flood
plain of the river to establish their relationship with the incidence of flooding and
inundation on the stated day. Alongside, the perception of the local people was
collected to validate the finding.
While climatic factors control the inputs that set the processes for genesis of urban
flooding, the physical changes in the urban landscape particularly those relating to
land use and drainage system largely modulate the flow and flow rates to produce
flooding damages resulting from the flood events caused to livelihoods of urban
dwellers. The responsibility of mitigation and adaptation falls in the domain of local
and central government. Infrastructure solutions that are developed to mitigate the
impacts of the flood have negative consequences on the environment. As seen in most
of the cases, in an attempt to develop permanent solutions, the river environment
is changed. For example, solutions like building flood walls and developing road
corridors along the river changes the flow regime by decreasing the river’s right
of way. The major problem of urban areas is the increasingly impervious surface,
so the most important aspect of urban flood management is to look into rainwater
management which is not sufficiently addressed and the worst part is that the storm
water management does not even come in the purview of urban development in
Nepal.
References
Dewan, Tanvir H. 2015. Societal Impacts and Vulnerability to Floods in Bangladesh and Nepal.
Weather and Climate Extremes 7: 36–42.
Gorelick, Noel, Matt Hancher, Mike Dixon, Simon IIyushchenko, David Thau, and Rebecca Moore.
2017. Google Earth Engine: Planetary-Scale Geospatial Analysis for Everyone. Remote Sensing
of Environment 202: 18–27.
IPCC. 2014. Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III
to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing
Team, eds. R.K. Pachauri and L.A. Meyer], 151, Geneva, IPCC.
Khanal, Bina. 2018. Recurrent Water Induced Disaster from Urban Flooding and the Resulting
Livelihood Consequences: A Case Study of Dhobi Khola Catchment. M.Sc, Nepal Engineering
College-Center for Postgraduate Studies (nec-CPS), Pokhara University.
Mirza, Monirul. 2011. Climate Change, Flooding in South Asia and Implications. Regional
Environmental Change 11: 95–107.
Molden, Olivia, and Tyler McMahon. 2019. Nepali Times. 29, 11. Accessed November 30, 2019.
https://www.nepalitimes.com/here-now/in-nepals-weather-erratic-is-the-new-normal/.
NWCF. 2009. The Bagmati: Issues Challenges and Propspects. Kathmandu: Nepal Water
Conservation Foundation (NWCF).
Prajapati, Rajaram, Thapa Bhesh Raj, and Rocky Talch. 2018. What Flooded Bhaktapur.
Kathmandu: My republica.
Rijal, Sushila, Bhagawat Rimal, and Sean Sloan. 2018. Flood Hazard Mapping of a Rapidly
Urbanizing City in the Foothills (Birendranagar, Surkhet) of Nepal. Land 7 (2).
Sada, Rajesh. 2012. Hanumante River: Emerging Uses, Competition and Implication. Journal of
Science and Engineering 1.
Salike, Inu Pradhan, and Jiba Raj Pokharel. 2017. Impact of Urbanization and Climate Change on
Urban Flooding: A Case of the Kathmandu Valley. Journal of Natural Resources and Development
7: 56–66.
Wang, Xiaolan L., and Yang Feng . 2013. RHtestsV4 User Maual. User Maual. Toronto: Climate
Research Division Atmospheric Science and Technology Directorate Science and Technology
Branch, Environment Canada, July 20.
Wrobley, Madison. 2020. Groundwater Quality and Quantity Are Linked, and Neither Are
Improving. The Kathmandu Post Daily Newspaper , January 21.
Zhanag, Xuebin, and Feng Yang. 2004. RClimDex (1.0) User Manual. User Manual, ON: Climate
Research Branch, Environment Canada Downsview.
Urban Drainage Study for Gopalganj
Pourashava Considering Future Climate
Change Impacts
Faruque Abdullah, A. K. M. Saiful Islam, Afsara Tasnia,

G. M. Tarekul Islam, Sujit Kumar Bala, and Nahruma Mehzabeen Pieu
Abstract This study focuses on an investigation, through hydraulic modelling,

of the urban drainage systems in designing suitable drainage infrastructures for
Gopalganj Pourashava by considering the possible impact of climate change on
drainage. The daily rainfall data of Madaripur station from the Bangladesh Meteo-
rological Department (BMD) for the base period of 1987–2016 has been analysed to
prepare the IDF (Intensity–Duration–Frequency) curves for different return periods
by fitting Extreme Value Type-1 (Gumbel) Distribution. Projected rainfall data has
been obtained from 11 Regional Climate Models simulated considering high-end
scenarios (RCP8.5) over the CORDEX South Asia domain for the near future (2017–
2046). Design hyetographs were developed for 10 years 2-hour storm, which was
later applied to generate runoff using the SWMM model. Water levels obtained
from the frequency analysis at Atharobanki and the peak flow of Old Madhumati
at Haridaspur for a 20-year return period are then used as boundary conditions for
the model. Afterwards, natural canals which overtopped during the simulation have
been identified. Finally, five alternative design conditions based on node flooding
have been proposed to improve the urban drainage system. A comparison between
the present and future drainage scenarios under different climate conditions has also
been represented.
Keywords Climate change · Gopalganj · Hyetograph · SWMM · Urban drainage
1 Introduction
1.1 Background of the Study
The planet’s hydrological cycle is likely to face severe impacts of climate change
and global warming due to the enhanced effects of greenhouse gas (IPCC 2014).
F. Abdullah (B) · A. K. M. S. Islam · A. Tasnia · G. M. T. Islam · S. K. Bala · N. M. Pieu

Institute of Water and Flood Management (IWFM), Bangladesh University of Engineering and
Technology (BUET), Dhaka, Bangladesh

24 F. Abdullah et al.
The average annual runoff will increase in high latitudes, in equatorial Africa and
Asia, and Southeast Asia, and will decrease in mid-latitudes and most subtropical
regions (Arnell 1999). The Ganges–Brahmaputra-Meghna (GBM) basins are one
of the world’s most significant and complex river nexuses. The water resources of
these river systems are highly vulnerable to global climate change (Solomon et al.
2007), which may have substantial insinuations on the livelihoods and well-being
of the people in the region ( Eriksson et al. 2009). Bangladesh is at the frontline of
impending climate change impacts. It is one of the most densely populated coun-
tries in the world with a small geographic extent and is located in the deltaic setting
of the GBM basins. This unique geographical location and proximity to the sea
and monsoonal climate have made it more vulnerable to increasing natural disas-
ters (Mondal et al. 2018). Varying levels of physical exposure, like storm surge,
flooding, and changing weather patterns, exacerbated by sea-level rise, represent
significant social and legal risks to its populations. Urban encroachment, rising sea
levels, and increased high-frequency storm surges are expected to intensify urban
drainage runoff impacting present infrastructures. An effective urban drainage design
reduces the potential impact of new and existing developments concerning surface
water drainage discharge. According to the 2011 population census of Bangladesh
Bureau of Statistics, the urban population was about 27% of the total population
in Bangladesh. Since independence, the average rate of urbanization is about 5%
(Ahmed and Ahmed 2012), and the percentage share of the urban population has
doubled, from 15% in 1974 to 28.4% in 2011 (BBS 2012). This extensive expansion
of urban areas as well as its economy would need urban planning. The Pourashavas,
which are district towns, would need to provide urban services to their citizens and
provide a sustainable environment for sound living.
Rapid urbanization and substantial increase in built-up areas in Gopalganj
Paurashava have taken place over the last few decades. These phenomena have facili-
tated a significant growth of impervious regions, hampered natural drainage patterns,
and reduced detention basins, leading to a lower time of concentration and increased
stormwater peak flow. Improper drainage systems very often lead to flooding, unsafe
and unhygienic conditions for humans and animals, and damaging numerous estab-
lishments. Unfortunately, toxic substances, such as waste from markets, poly bags,
household wastewater, fertilizer, and other chemicals, are regularly drained into
stormwater drainage systems. These toxic chemicals lead to water pollution, proving
unhealthy for fish, plants, and different water life, even killing them. Therefore, it
is essential to plan for a coordinated stormwater drainage management system for
the proficient regulation of stormwater quantity. The initial step for this management
planning is to conduct a drainage analysis via computer modelling, which will be
efficient in prescribing fitting solutions to manage drainage problems better.
In this study, EPA’s Storm Water Management Model (SWMM), the most widely
recognized model for urban drainage as well as sanitary sewer systems, is used for the
assessment of present and future drainage capacity in Gopalganj Pourashava as part
of the drainage and environmental management master plan under the Municipal
Governance and Services Project (MGSP) of the Local Government Engineering
Department (LGED). SWMM is unique because of its excellent ability to simulate
Urban Drainage Study for Gopalganj Pourashava … 25
hydraulics and hydrology within the same interface compared to other commonly
used models. The advantage of SWMM over other models is that it allows the users
to prepare the inputs manually and run the model smoothly, thereby providing users
access to all the functions and tools to create and edit data. EPASWMM 5.0 is
a dynamic rainfall-runoff simulation model used for a single event or long-term
assessment of runoff quantity and quality from urban areas (Rossman and Huber
2016). SWMM has been used in numerous sewer and stormwater studies throughout
the world. It can design and size drainage system components for flood control, flood
plain mapping of natural channel systems, develop control strategies for minimizing
combined sewer overflows, and evaluate the impact of inflow and infiltration on
sanitary sewer overflows, etc.
The Institute of Water and Flood Management (IWFM) of Bangladesh Univer-
sity of Engineering and Technology (BUET) has used SWMM for several studies
relating to urban runoff and drainage (Khan and Chowdhury 1998; Khan et al. 2006;
Chowdhury et al. 1998; Rahman et al. 1999; Asian Development Bank 2013). (Akter
and Tanim 2016) used the SWMM model and spatial analysis in ArcGIS to iden-
tify urban-flood-prone areas in Chattogram that were validated using recent field
studies. This study provided insights for flood management-related decision-making
approaches to identify and emphasize the vulnerable flood-prone zones in the study
area. Another study conducted by (Tarek et al. 2017) on the Chaktai Khal basin area
in Chattogram used GeoSWMM to simulate and analyse the behaviour of flash floods
in the watershed. Their analysis revealed that a maximum of 47% of the total study
area is prone to flash floods during the rainy season, suggesting that SWMM can
effectively analyse integrated flash floods in an ungauged urban system. According
to (Cambez et al. 2008), SWMM can also produce a satisfactory outcome in the
long-term modelling of a metropolitan area despite having any minor limitations in
catchment hydrological description. Long-term simulations help compare different
storage scenarios and sewage treatment plant capacities. (Hossain et al. 2019) indi-
cated these limitations and concluded that event-based modelling of EPASWMM
outperforms the continuous simulation approach in evaluating both direct runoff
hydrograph and total runoff hydrograph. (Paul et al. 2013) led a study in three coastal
towns—Amtali, Galachipa, and Pirojpur and used SWMM to evaluate the future
drainage condition of these areas. Their research revealed that rapid urban develop-
ment in coastal zones and climate change-induced sea-level rise would significantly
increase inundation vulnerability compared to the baseline period. (Bai et al. 2018)
used SWMM to assess the LID (Low Impact Development), a storm management
technique at Sucheng district in China, under four different scenarios. The study indi-
cates that runoff frequency reduces with increasing rainfall under all scenarios until it
becomes stable with a certain rainfall amount. (Khadka and Basnet 2019) have found
SWMM as an effective modelling tool for managing storm waters in major cities of
Nepal, where the management of overflow from the existing drainage is a crucial
issue. The study was conducted on the Barahi Chowk area of Lakeside, Pokhara,
focusing on stormwater drainage design and also a comparison with the existing
drainage system. Another application of SWMM is to estimate urban runoff peaks
and volume. For example, a study done by (Rabori and Ghazavi 2018) assessed urban
flooding of a semi-arid area such as Zanjan city of Iran, concluding that this model
can simulate urban peak flow with acceptable accuracy. (Jang et al. 2007) suggested
SWMM for evaluating pre and post-development conditions of a metropolitan area
to study hydrologic impact assessments. However, employing a synthetic hydro-
graph for pre-development and an urban hydrology model for post-development
may introduce some unavoidable difficulties which can be resolved using a single
SWMM model. The natural streams are converted to artificial drainage networks
due to urbanization-induced land use change (Haase 2009). This phenomenon ulti-
mately results in urban flooding in monsoon regarding high slopes and scarcity of
a proper drainage management system. In this connection, the assessment of the
rainfall-runoff process is essential, and (Babaei et al. 2018) used SWMM, which
showed good accuracy for peak runoff simulation for Urmia city of Iran.
The prime goal of the study was to investigate the detailed hydrological and
hydraulic analysis of the urban watersheds in determining suitable drainage systems
for the study area incorporating the probable impact of climate change on drainage.
Secondary objectives include hydraulic design and sizing of the primary and
secondary canals for flood control, sizing detention facilities, and designing control
strategies to minimize sewer overflows. SWMM was set up, tested, and simulated
for present conditions and the near future (2017–2046). Impacts of climate change
on the urban drainage system have also been considered to design climate-resilient
urban infrastructures. Finally, a set of alternative solutions to the existing and future
drainage issues have been recommended to improve the urban drainage capacity of
Gopalganj Pourashava.
1.2 Study Area and Existing Drainage System
Gopalganj district, located at 23°000 47.67 N 89°049 21.41 E, stands on the bank
of the Madhumati River. It is bounded by Faridpur district in the North, Pirojpur and
Bagerhat in the South, Madaripur and Barisal in the East, and Narial in the West.
Gopalganj Pourashava town is the district town of Gopalganj District that lies within
the Sadar Upazila. After the liberation war, Gopalganj Sadar becomes Pourashava,
consisting of 9 wards and 75 mahallas. The Local Government Engineering Depart-
ment (LGED) has formulated a master plan for developing a practical and affordable
solution to the stormwater drainage problem of Gopalganj Pourashava with an exten-
sion area that covers a total area of 35.32 sq. km (Fig. 1). The study area bears the
characteristics of rapid urbanization, leading to a gradual encroachment of the natural
canals and hindrance to the smooth flow of drainage water and thus has become a
matter of suffering for the people of the Pourashava.
The natural canal system of the study area comprises four significant channels.
The Old Madhumati River, now known as the Mora Modhumati, is the core natural
drainage canal flowing through the centre of the town. The other three significant
canals known as ‘Khals’ within the area have been originated from the Old Madhu-
mati. But the gradual encroachment and unplanned infrastructures result in unstable
Fig. 1 Study area and existing drainage condition of the Pourashava
flow conditions, allowing the river to gain silt in some locations and hampering the
natural flow from upstream to downstream. In this connection, discharge capacity
has been decreased to a great extent which is the main factor behind the drainage
congestion and waterlogging in the city area. Among the other three channels, Panshi
Khal, which once actively carried flow from the Old Madhumati to the downstream,
is now almost dry with a minimum flow circulation. Boiragi Khal is the most impor-
tant of the three because it has the largest drainage outfalls from the market area.
But the entrance from Madhumati is now totally blocked, and no water can enter
Boiragi Khal from the river. So, to improve the present condition, the connectivity
of the natural canals should be given first priority before talking about the artifi-
cial drainage conduits. Another disengaged channel, known as Pachuria Khal, meets
Boiragi Khal and then carries the flow to the Bonnir Baor, which is considered one of
the most significant outfalls for the entire catchment. The efficiency of the existing
primary drainage network is also not satisfactory. The drainage channels got reduced
both by horizontal and vertical dimensions in the city area. Once plied upon local
launches and country boats, encroachment and siltation have turned these large canals
into a non-distinguishable small drain. At present, the drain is clogged with market
garbage, water hyacinth, etc., creating blockages in several points. These locations
are to be opened up again, in order to make the drain more effective and drain water
towards the New Madhumati River.
2 Methodology
Several steps have been followed for the urban drainage modelling of the Gopalganj
Pourashava. The analysis of rainfall data from available rain gauges involves sepa-
ration of rainfall events, estimating the return periods of a certain durational rainfall,
evaluating data inconsistency, etc. The stormwater model was developed using time
series of daily and sub-daily rainfall events to determine runoff. Sub-catchments are
hydrologic land units with dynamic terrain and drainage characteristics that route
surface runoff to a single outlet point. These are delineated using spatial analysis
of the existing digital elevation model or topological survey data. By utilizing the
study area’s land use map, sub-catchments are divided into permeable and imper-
meable subareas. Drainage systems have been mapped using a Digital Elevation
Model (DEM) that was trained by existing drainage networks and then subjected to
spatial analysis. The cross-sectional survey of the principal channels and drainage
structures yielded information on the prevailing drainage characteristics of the area.
Finally, a number of input parameters, such as soil characteristics, infiltration rate,
manning’s coefficient, etc., have been used based on field observation and literature
review. The model has been simulated using the design storm scenarios. A hypothet-
ical rainstorm event with a defined duration, temporal distribution, rainfall intensity,
return period, and total depth of rainfall is referred to as a design storm. To analyse
future climate change vulnerabilities, the SWMM model has been used to replicate
design storm scenarios with and without climate change impacts. Statistical anal-
ysis has also been conducted to determine the design water level of the surrounding
rivers used as boundary conditions of the model. Based on the design simulation
results, the drainage network capacity has been assessed, and various alternative
options have been suggested to improve the performance of the drainage networks.
A similar exercise has been conducted for the future periods considering climate
change where both the shift in land use pattern and evolution of rainfall and water
levels of the surrounding rivers for future periods are considered for simulating the
model.
2.1 Data Collection and Watershed Delineation
The combination of EPA’s Storm Water Management Model (SWMM5) and

GeoSWMM have been used in this study to conduct urban drainage modelling. For
the stormwater drainage analysis, collecting and compiling all hydraulic and morpho-
logical data concerning the urban drainage systems and the surrounding rivers have
been ensured. For the watershed delineation, SRTM 1 Arc-Second Global DEM
of 30 m resolution was collected from the USGS Global Data Explorer website. In
addition, the historical daily rainfall data was collected from Bangladesh Meteorolog-
ical Department (BMD), and the future projected rainfall information using regional
climate models over the project area has been obtained from Mohammed et al. (2018).
The required topographic data and bathymetry of the corresponding canals were
collected through a field survey. Moreover, necessary information regarding existing
drainage canals, cross drainage structures, bridges and culverts, and the projected
land use change of the study area has been collected from the Local Government
Engineering Department (LGED) and Gopalganj Pourashava office. The watershed
delineation tool of GeoSWMM, the advanced version of EPA’s SWMM5, has been
used to delineate the sub-catchments of the study area. The process involves DEM
clipping, stream burning, stream and outlet generation, watershed delineation, etc., as
GeoSWMM facilitates users to conduct these operations using the ArcGIS interface
directly. After delineating the sub-catchments and assigning the necessary properties
for the corresponding streams, joints, and links, the whole project was then imported
to the EPASWMM5 interface for the model development, simulation, and urban
drainage modelling.
The drainage model has one upstream boundary and three downstream bound-
aries shown in Fig. 2(c). The upstream boundary is at the mouth of the Old Madhu-
mati River, where it receives flow from the Madaripur Beel Route. The downstream
boundary of the Old Madhumati River is at the confluence of the Madhumati River.
Fig. 2 a Delineated watersheds within the extension area. b streams, links, and joints within the
watersheds. c locations of the inflow and stage boundaries for simulating SWMM
The downstream boundaries of Panshi Khal and Pachuria Khal have been considered
as open water bodies.
2.2 Hydrologic Analysis
The hydrologic analysis mainly focuses on predicting the design water level and
outflow through frequency analysis. The primary method in frequency analysis
involves fitting various probability distribution functions to historical yearly peak
water levels and discharge data at appropriate locations for the selected design return
period (Paul et al. 2013). The design return period has been considered based on
the size and importance of the proposed intervention. The gauge stations main-
tained by BWDB nearest to the Gopalganj Municipality are located at Haridaspur
of Madaripur Beel Route and the off-take at Atharobanki of the Gorai-Madhumati-
Haringhata-Baleswar River. Thus the water level data of Madaripur Beel Route at
Haridaspur and Gorai-Madhumati-Haringhata-Baleswar River at Atharobanki have
been used for frequency analysis. The discharge data of the Madaripur Beel Route at
Haridaspur has been taken from (Hossain and Chowdhury 2018) where the highest
recorded historical peak flow was 1550 m3 /s in 1998.
With regard to frequency analysis, five probability distribution functions (PDFs)
were used. These were Pearson Type III (P3), Log Pearson Type III (LP3), Two-
Parameter Log-Normal (LN2), Three-Parameter Log-Normal (LN3), and Gumbel
(EV1) distribution. A probability plot correlation coefficient (PPCC) was used to
test them later (Filliben 1975). The best-fitted PDFs are P3 for the annual maximum
water level of Gorai-Madhumati-Haringhata-Baleswar River at Atharobanki (Table
1(b)) and Madaripur Beel Route at Haridaspur (Table 1(c)) and LN3 for the peak
flow of the Madaripur Beel Route at Haridaspur (Table 1(a)).
For the yearly maximum water level and peak flow of the Madaripur Beel Route
at Haridaspur and water level of Gorai-Madhumati-Haringhata-Baleswar River at
Atharobanki, Fig. 3 illustrates probability diagrams with a 90% confidence interval.
The observed values for the annual maximum water level of the Madaripur Beel Route
at Haridaspur and the Gorai-Madhumati-Haringhata-Baleswar river at Atharobanki,
as well as the yearly peak flow of the Madaripur Beel Route at Haridaspur, fit
well within the 90% confidence interval of the adjusted P3 and LN3 distributions,
respectively.
Following the importance and size of the proposed drainage master plan, a 20-year
return period has been considered as the design return period as per the recommen-
dations of Local Government Engineering Department. The design water levels of
Madaripur Beel Route at Haridaspur and Gorai-Madhumati-Haringhata-Baleswar
River at Atharobanki corresponding to a 20-year return period are 4.65 m and
4.07 m, respectively. The design peak flow of Madaripur Beel Route at Haridaspur
corresponding to a 20-year return period is 1412 m3 /s.
Table 1 a Adjusted PDFs and associated PPCC values for annual maximum flow (m3 /s) of
Madaripur Beel Route at Haridaspur. b Adjusted PDFs and associated PPCC values for annual
maximum water level (m PWD) of the Gorai-Madhumati-Haringhata-Baleswar River at the off-
take at Atharobanki. c Adjusted PDFs and associated PPCC values for annual maximum water level
(m PWD) of Madaripur Beel Route at Haridaspur
PDFs Return period PPCC Rank
2.33 20 50 100
(a) Annual peak flow (m3/s)
LN2 782 1302 1503 1652 0.94394 5
LN3 731 1412 1792 2120 0.95838 1
P3 765 1398 1651 1840 0.95817 2
LP3 736 1414 1797 2138 0.95751 3
EV1 786 1374 1603 1774 0.94434 4
(b) Annual maximum water level (m PWD)
LN2 3.07 4.79 5.43 5.89 0.90034 2
LN3 3.19 4.05 4.27 4.42 0.90027 3
P3 3.17 4.07 4.32 4.49 0.90211 1
LP3 3.34 3.77 3.78 3.78 0.82954 5
EV1 3.05 4.22 4.67 5.01 0.89983 4
(c) Annual maximum water level (m PWD)
LN2 3.37 4.87 5.40 5.78 0.97838 4
LN3 3.50 4.62 4.94 5.15 0.98266 2
P3 3.48 4.65 4.99 5.22 0.98313 1
LP3 3.45 4.64 4.96 5.17 0.98110 3
EV1 3.37 4.78 5.33 5.74 0.97219 5
Fig. 3 Probability distribution diagram with 90% confidence interval of the a P3 distribution fitted
to the annual maximum water level of Madaripur Beel Route at Haridaspur. b P3 distribution fitted
to the annual maximum water level of the Gorai-Madhumati-Haringhata-Baleswar River at the off-
take of Atharobanki. c LN3 distribution fitted to the annual peak flow of Madaripur Beel Route at
Haridaspur
2.3 Evaluation of Climate Change Impacts on Boundary

Flows
The Madaripur Beel Route has a connection with the Padma River, which receives
flow both from the Ganges and Brahmaputra. Changes in future flow of the Ganges
and the Brahmaputra due to climate change will influence the variations in the flows
of the Madaripur Beel Route and other surrounding rivers. At IWFM, BUET, a semi-
distributed hydrological model, SWAT has been applied to assess climate change’s
impact on the major river flow (Mohammed et al. 2017) under the HELIX research
project. The Soil and Water Assessment Tool, SWAT (Arnold et al. 2012), was used
to simulate future discharges with 11 different climate projections, as shown in Table
2.
Using the boundary discharge from the SWAT model, a one-dimensional hydro-
dynamic model HEC-RAS has been simulated to determine the changes of flow in the
major rivers of Bangladesh (Mondal et al. 2018). The downstream boundary of the
HEC-RAS was obtained from another three-dimensional modelling suite, Delft3D,
which has incorporated the impact of sea-level rise from its open boundary condi-
tions. In the fifth assessment report of IPCC (IPCC 2014), a global mean sea-level
rise of 52–98 cm by the year 2100 under a very high emission scenario of RCP 8.5
and 36–71 cm under a moderate emission scenario of RCP 4.5 have been projected.
For the Bay of Bengal, the IPCC projection is a little more comprehensive (25–73 cm
under RCP 4.5). Based on IPCC’s projections of global and local (Bay of Bengal)
sea-level rise, a maximum sea-level rise of 1 m by 2100 and 0.31 m by 2050 has been
anticipated for the Bay of Bengal. The downstream sea boundary of the Delft3D based
coastal model has been modified according to these projections of sea-level rise. The
HEC-RAS model has been rerun for flood discharges of 100 year return period to
Table 2 List of Regional Climate Models used in this study

Sl Institute GCM RCM Res RCP
1 CSIRO ACCESS1.0 CCAM-1391 M 0.5° 8.5
2 CSIRO CCSM4.0 CCAM-1391 M 0.5° 8.5
3 SMHI CNRM-CERFACS-CNRM-CM5 RCA4 0.5° 8.5
4 CSIRO CNRM-CM5 CCAM-1391 M 0.5° 8.5
5 SMHI ICHEC-EC-EARTH RCA4 0.5° 8.5
6 CSIRO MPI-ESM-LR CCAM-1391 M 0.5° 8.5
7 MPI-CSC MPI-M-MPI-ESM-LR REMO2009 0.5° 8.5
8 SMHI MPI-M-MPI-ESM-LR RCA4 0.5° 8.5
9 SMHI NOAA-GFDL-GFDL-ESM2M RCA4 0.5° 8.5
10 SMHI IPSL-CM5A-MR RCA4 0.5° 8.5
11 SMHI MIROC-MIROC5 RCA4 0.5° 8.5
generate stage hydrographs along the major rivers of the basin, and a significant
increment in flood peaks has been observed due to climate change impacts.
Based on the SWAT simulation for baseline (1980–2009) and future (2070–2099)
and the HEC-RAS model simulation from the studies mentioned in the previous
paragraph, the anticipated changes in flow and water level due to climate change
and sea-level rise impact have been incorporated with the urban drainage design
analysis. As Madaripur Beel Route is a branch of the Arial Khan River, the evolution
of discharge in Mawa has been taken as 6% by the end of the century, according to
(Mondal et al. 2018). Similarly, variation in the Ganges’ discharge at Gorai off-take is
29%, and the Brahmaputra at Dhaleswari is 14% by the end of this century resulting
from climate change impacts. Considering this deviation, the possible increase in
the peak discharge in Madaripur Beel Route at Haridaspur will be about 10% (one-
third of the change in Ganges flow) by 2030 according to the study. Again, the peak
water level difference due to sea-level rise in the Ganges at Gorai is about 62 cm,
the Brahmaputra at Dhaleswari is 38 cm, and the Ganges at Mawa is about 54 cm by
2100. In this connection, the peak water level at Atharobanki has been considered to
increase by about 21 cm (one-third of the change in the Ganges at Gorai for 2100)
in the near future (2017–2046).
2.4 Rainfall Analysis
The rainfall analysis includes developing Intensity–Duration–Frequency (IDF)

curves through the frequency analysis of rainfall data. This operation is followed
by the generation of hyetographs used for rainfall-runoff analysis to estimate peak
runoff rate. For Gopalganj Pourashava, daily rainfall data from 1977 to 2016 at the
nearest station Madaripur regarding the present scenario has been collected from
Bangladesh Meteorological Department (BMD). This data has been analysed and
used for developing IDF curves for long and short duration rainfall and eventually
hyetographs for the base period of 1987–2016. As climate change will influence the
future precipitation trend, the rainfall data has also been projected for the Madaripur
station for the near future (2017–2046). Here, eleven CMIP5 regional climate models
have been used to produce a monthly delta. The maximum monthly value of 1-day
rainfall (RX1) of each year of the baseline period (1987–2016) was calculated for
each model. A similar analysis has been done for the near future (2017–2046). The
percent differences between the averages of these two data sets indicate a delta for
that particular month. A similar calculation was carried out for each month from April
to October. These delta values were then used to project the future rainfall from the
observed rainfall data set and develop consequent IDF curves and hyetographs. The
Extreme Value Type I (Gumbel) Distribution has been used for creating the curves for
different return periods. For generating the IDF curves regarding short-duration rain-
fall, the IDF curves for long-duration were plotted in log–log scale, which resulted
in almost linear graphs. From the equations of those trend lines, rainfall intensities
were found for different durations and return periods. The duration step was 10 min.
Fig. 4 a Long duration IDF curves. b short duration IDF curves for different return periods. c
design hyetographs for 10-year 2-h rainfall for the baseline (1987–2016). d Long duration IDF
curves. e short duration IDF curves for different return periods. f design hyetographs for 10-year
2-h rainfall for the near future (2017–2046)
Using these short-duration IDF curves, design hyetographs were then developed both
for the present (1987–2016) and future climate change (2017–2046) scenarios using
the alternating block method. A series of hyetographs were constructed for return
periods of 2, 5, 10, 25, 50, and 100 years, with the 10-year 2-h hyetograph being
chosen as the study’s design hyetograph (Fig. 4).
3 Results and Discussions
3.1 Canal Capacity Assessment under Different Simulation

Scenarios
SWMM model has been set up and simulated for the normal flood condition to assess
the existing canal capacity. The model has also been run for the design condition,
and finally considering the future climate change impacts to evaluate the necessity
of designing a suitable and efficient urban drainage system for the study area. Due to
the lack of observed water level and discharge data in the Old Madhumati River and
other surrounding major canals, it was not possible to compare the model flow with
the historical data. In the absence of the flow data in the channels, the model has been
set up using insights from the drainage modelling of adjacent coastal cities used in
previous studies (Paul et al. 2013; Hossain and Chowdhury 2018). In addition, the
primary calibration parameter, such as Manning’s roughness coefficient, has been
used in this study considering the soil type, land use pattern, and insights from a
similar project in the nearby region conducted by IWFM (Asian Development Bank
2013). Table 3 summarizes the location and magnitude of all the inflow and stage
boundaries of the SWMM model for the simulations under regular flooding, design
flooding, and future climate change scenarios based on (Mondal et al. 2018; IWM
2011).
A model has been simulated for normal flood conditions (a flood with a 2.33 year
return period), a 10-year 2-h design storm hyetograph considering climate change
implications on precipitation, and also without considering climate change impacts.
The maximum amount of discharge entering the Old Madhumati River from the
Madhumati River is about 15%, based on the IWM study report. Therefore, the
upstream inflow for the normal flood condition to the Old Madhumati River from
the Madhumati River has been considered as 110 m3 /s which is 15% of the flow in
Madhumati River (731 m3 /s) shown in Table 1(a). Similarly for the design storm
condition with a 20-year return period, the design inflow to the Old Madhumati
River has been set as 212 m3 /s (15% of 1412 m3 /s). For the design storm condition
considering climate change impacts, an extra 10% increment to 212 m3 /s has been
considered regarding the high-end climate scenario. A similar convention has been
followed for the corresponding stage boundaries at Atharobanki and Haridaspur, as
shown in Table 1(b) and (c), respectively. Considering the impact of sea-level rise
in the downstream boundary, about 21 cm additional rise of water level has been
included with 4.07 m water level at Atharobanki for a 20-year return period based
on the outputs of (Mondal et al. 2018). For the stage boundaries at the outfall of
Pachuria and Panshi Khal, normal depth has been considered. Manning’s equation
has been modified and the normal depth has been calculated for both the locations
as 2.2 and 2.0 m, respectively.
Figure 5 indicates that the water surface has maintained a constant below 4 m
threshold for the Old Madhumati River under normal flood conditions, representing
similar findings compared to a study conducted by the Institute of Water Modelling
back in 2009. It has also been found that water surface height maintains a 3.4–2.9 m
range above the ground level in various canal sections in case of regular flooding.
Under the design flooding condition, the range is 4.3–2.9 m, and for the climate
change scenario, it is 4.4–2.9 m. The capacity of the drainage systems, including
major canals, has been assessed for this high tide and high-intensity rainfall events
for Gopalganj Pourashava at the normal flooding condition. Analysis of water surface
profiles indicates that, except for a few locations, the capacity of the canal system is
enough to drain the water during high tide. As ponding is allowed during the simu-
lation, the overtopping margin would approximate the quantity of water overflowed.
No significant overtopping has been observed along the canals under the present
condition. Therefore, the existing drainage system is capable of carrying the flow of
normal floods. Under design conditions, overtopping is noticed at six locations in
the Old Madhumati River and one location in Boiragi Khal. The simulation consid-
ering climate change exhibits significant overtopping at eight locations in the Old
Table 3 Location and magnitude of the inflow and stage boundaries of the SWMM model regarding
three simulation scenarios
Location Node Simulation Value Boundary type Remarks
Condition
Entrance of the J1 Normal 110 m3 /s Inflow 15% of the normal
Old Madhumati flood at Haridaspur
River in the Madhumati
River
Design 212 m3 /s 15% of the flow at
Haridaspur in
Madhumati River
for 20-year return
period flood
Climate 233 m3 /s 15% of the design
Change flood at Haridaspur
in the Madhumati
River for 20-year
return period flood
and additional 10%
increment
considering climate
change
Outfall of the Old O1 Normal 3.17 m Stage Normal flood stage
Madhumati River at Atharobanki
station in the
Madhumati River
Design 4.07 m Stage at
Atharobanki station
on Madhumati River
for 20-year return
period flood
Climate 4.28 m Stage at
Change Atharobanki station
in the Madhumati
River considering
20-year return
period flood and
additional 21 cm for
climate change
impact due to
sea-level rise
Outfall of O2 Normal 2.2 m Stage Normal depth
Pachuria Khal Design 2.2 m Normal depth
Climate 2.2 m Normal depth
Change
Outfall of Panshi O3 Normal 2.0 m Stage Normal depth
Khal Design 2.0 m Normal depth
(continued)
Table 3 (continued)
Location Node Simulation Value Boundary type Remarks
Condition
Climate 2.0 m Normal depth
Change
Fig. 5 Location of the nodes flooded and corresponding water surface profile of the Old Madhumati
River under a normal flooding condition. b design flooding condition. c future climate change
scenario
Madhumati River and Boiragi Khal. In general, the existing urban drainage system
of the Pourashava is not capable of carrying the flow during the design condition and
considering future climate change and sea-level rise impacts.
3.2 Alternative Design Options
SWMM simulation indicates that the drainage system of the study area will be
suffering from urban flooding in the future. The drainage systems’ present capacity,
including the Old Madhumati River and other canals, is insufficient to drain the
Fig. 6 Location of the nodes flooded under six alternative design options considering future climate
change impacts
runoff generated from the design storm and climate change impacts on the rainfall,
boundary inform, and sea-level rise. SWMM model has been simulated considering
six alternative design scenarios, and Fig. 6 represents the location of the nodes flooded
under these options, considering climate change scenarios. These alternative design
options are
• Option 1: Establishing a connection between the Old Madhumati River and
Boiragi Khal.
• Option 2: Establishing a connection between the Old Madhumati River and both
Pachuria Khal and Boiragi Khal.
• Option 3: Dredging of the Old Madhumati River considering vertical bank
protection using retention walls (rectangular channel).
• Option 4: Dredging of the Old Madhumati River considering vertical bank protec-
tion using retention walls (rectangular channel) and connecting Boiragi Khal with
the Old Madhumati River.
• Option 5: Dredging of the Old Madhumati River and Boiragi Khal considering
trapezoidal channel with bank protection and connecting Boiragi Khal with the
Old Madhumati River.
• Option 6: Dredging of the Old Madhumati River, Boiragi Khal and Pachuria Khal
considering trapezoidal channel section and connecting both Boiragi Khal and
Pachuria Khal with the Old Madhumati River.
A summary of different options used to improve the urban drainage system and the
comparison of node flooding for existing conditions and alternatives are presented in
Table 4. The least amount of node flooding is observed in the Old Madhumati River
under Option 6. Considering the increase of the freshwater flow in Boiragi Khal,
which will eventually reduce deposition of sediments and improve water quality,
Option 6 has appeared to be more ecosystem friendly and less costly, in comparison
with other options. However, dredging of Old Madhumati River, Boiragi Khal, and
Pachuria Khal is essential in order to improve the overall carrying capacity.
It has been observed that urban runoff is wholly carried out by the improved
trapezoidal section, and has established full connectivity of these surrounding Khals
with the Old Madhumati River. Consequently, the overtopping is noticed only in a
Table 4 A comparison of simulations between existing conditions and alternative options for
improving the performance of urban drainage system
Option Old Madhumati Boiragi Pachuria
Existing Connection with old Madhumati NO NO
—
X-Section Irregular Irregular Irregular
Nodes flooded 7 1 0
1 Connection with old Madhumati YES NO
—
Nodes flooded 7 4 0
2 Connection with old Madhumati YES YES
—
Nodes flooded 5 3 2
3 Connection with old Madhumati NO NO
—
X-Section Rectangular Irregular Irregular
Nodes flooded 3 1 0
—
X-Section Rectangular Irregular Irregular
Nodes flooded 3 2 0
—
X-Section Trapezoidal Trapezoidal Irregular
Nodes flooded 3 2 0
6 Connection with old Madhumati YES YES
—
X-Section Trapezoidal Trapezoidal Trapezoidal
Nodes flooded 1 0 0
few locations through the whole canal system of the study area. Hence, Option 6 is
recommended for the improvement of the drainage system of Gopalganj Pourashava.
3.3 Limitations of the Study
The following are some of the study’s limitations:

• The proper catchment delineation of the study region necessitates the use of
high-resolution DEM data. To build a high-resolution DEM for the study area, a
comprehensive survey using ground stations or the purchase of high-resolution
satellite images is required.
• The model could not be directly calibrated and tested due to a lack of site-specific
observed data. However, it has been evaluated using secondary data from prior
IWM and BWDB investigations. So it is really important to take necessary actions
to solve the data scarcity problems of the study area for future research work. Water
level from various significant canals and rainfall data in the research area should
be collected continuously during the monsoon season for a number of storms. The
major canals must be equipped with automatic water level and flow measurement
devices, which should be kept operational throughout the season. Automated rain
gauges can also be installed to record hourly rainfall data during storms.
4 Conclusion
In this study, the open source and community model SWMM developed by the EPA
was used to analyse urban drainage system of Gopalganj Pourashava. Due to the
lack of observed flow data in the canals, the model has been validated with similar
study results in this region. Models seem reasonably adequate to capture the shape of
the water surface profile of channels and stages of the surrounding rivers for normal
flood conditions. The urban drainage system of the study area is already facing several
challenges from natural hazards and human activities such as colossal sedimentation,
illegal encroachments, instability of banks, poor connectivity of sanitary sewerage,
solid waste disposal, etc. Again, various sections of the natural canals are overtopped
for the design condition. Climate change impacts on the urban drainage systems of
Gopalganj Pourashava would also cause overtopping and urban flooding in the Old
Madhumati River, Boiragi Khal, and Pachuria Khal. It is essential to take necessary
steps to combat urban flooding due to overflow depending on the importance of the
location, cost, and land availability. In this study, some options to improve the urban
drainage network have been considered. The best choice to resolve the drainage
congestion is to dredge the Old Madhumati River, Boiragi Khal, and Pachuria Khal,
considering a trapezoidal channel section with side slope 1:1.5 and connecting both
Boiragi Khal & Pachuria Khal with the Old Madhumati River.
Based on the modelling study and field visits, a number of challenges were found,
which will cause severe urban flooding in Gopalganj Pourashava. A summary of the
recommendation to improve urban drainage systems has been presented as follows:
• The entrance of the Old Madhumati River is partially opened to avoid deposition
of huge sediment load in the river, and dredging was found essential. This will
bring more freshwater discharge into the Old Madhumati River from Madhumati
River.
• The Old Madhumati River and other canals have been encroached upon illegally.
Illegal possessions on both sides of canals should be re-occupied, and the canal’s
waterways should be marked.
• Because of the steep slope (1:1) and illegal pipe connections, the current bank
protection of the canals is insecure in several areas. It should be designed in
accordance with LGED standards for bank slopes (1:1.5), construction materials,
compaction, and layouts.
• Household and commercial waste, along with sanitary sewage, are currently
disposed of directly into the stormwater drainage system. To improve the water
quality and reduce sedimentation in the canal, a separate sanitary sewage system
is required. Before draining to canals and natural water bodies, wastewater
from business and residential housing compounds should be treated at a sewage
treatment plant (STP).
• Pipe drains have been found to work well; they are environmentally benign and
carry less solid waste. When the discharge is high (>1.62 m3 /s), pipe drains with
proper manhole spacing are highly recommended.
• Construction of two or more primary and secondary canals, rather than one
tertiary-secondary canal, is advised when peak runoff exceeds the maximum
capacity of the pipe drain connecting to the other side of the natural drainage
system (e.g. Madhumati River).
• Solid wastes are currently disposed of in canals and river systems, reducing
the canal’s carrying capacity and degrading the ecosystem. In the Pourashava,
solid waste management systems should be enhanced by scheduling collection,
providing enough garbage bins, raising public awareness, and developing envi-
ronmentally safe dumping locations. Again, the Department of Environment’s
Bangladesh Standards and Guidelines for Sludge Management (2015) should
be followed while dealing with sludge from municipal and domestic sewage
treatment plants (STP).
• Both Boiragi Khal and Panshi Khal are now cut off from the Old Madhumati River,
causing increased sedimentation in the canal and environmental deterioration.
These canals should be connected to the Old Madhumati by building appropriate
water infrastructures such as bridges and culverts.
• Since they help to reduce peak urban floods ditches, ponds, and wetlands should
all be identified, conserved, and excavated (if necessary).
Acknowledgements The authors would like to thank the Institute of Water and Flood Manage-
ment (IWFM), BUET, who have provided the system components and support during the project
execution. The authors would also like to thank the Local Government Engineering Department
(LGED) for the financial support under the Municipal Governance and Services Project (MGSP).
References
Ahmed, S.S., and Muntasir Ahmed. 2012. Urbanization and Economic Development of Bangladesh:
The Primacy of Dhaka and Competitiveness. 2009: 1–21.
Akter, Aysha, and Ahad Hasan Tanim. 2016. Estimating Urban Flood Hazard Zones Using Swmm
in Chittagong City. Technical Journal River Research Institute 13 (1): 87–101.
Arnell, Nigel W. 1999. Climate Change and Global Water Resources. Journal of Global
Environmental Change, Volume 9, Supplement 1. https://doi.org/10.1016/S0959-3780(99)000
17-5.
Arnold, J.G., D.N. Moriasi, P.W. Gassman, K.C. Abbaspour, M.J. White, R. Srinivasan, C. Santhi,
et al. 2012. SWAT: Model Use, Calibration, and Validation. Transactions of the ASABE 55 (4):
1491–1508.
Asian Development Bank. 2013. People’s Republic of Bangladesh: Strengthening the Resilience of
the Urban Water Supply, Drainage, and Sanitation to Climate Change in Coastal Towns (Financed
by the Japan Fund for Poverty Reduction), October: 596.
Babaei, Sahar, Reza Ghazavi, and Mahdi Erfanian. 2018. Urban Flood Simulation and Prioritization
of Critical Urban Sub-Catchments Using SWMM Model and PROMETHEE II Approach. Physics
and Chemistry of the Earth 105 (February): 3–11. https://doi.org/10.1016/j.pce.2018.02.002.
Bai, Yiran, Na Zhao, Ruoyu Zhang, and Xiaofan Zeng. 2018. Storm Water Management of Low
Impact Development in Urban Areas Based on SWMM. Water (Switzerland) 11 (1). https://doi.
org/10.3390/w11010033.
BBS. 2012. National Population and Housing Census 2011. National Report. Government of
Bangladesh, National Planning Commission Secretariat, Bangladesh Bureau of Statistics.
Cambez, M.J., J. Pinho, and L.M. David. 2008. Using SWMM 5 in the Continuous Modelling of
Stormwater Hydraulics and Quality, 1–10.
Chowdhury, J.U., Md. Rezaur Rahman, Sujit Kumar Bala, and A.K.M. Saiful Islam. 1998. Impact of
1998 Flood on Dhaka City and Performances of Flood Control Works. Technical Report. IWFM,
BUET, Dhaka.
Eriksson, Mats, Xu Jianchu, Arun Bhakta Shrestha, Ramesh Ananda Vaidya, Santosh Nepal, and
Klas Sandström. 2009. “The Changing Himalayas: Impact of climate change on water resources
and livelihoods in the greater Himalayas.” Journal of Geophysical Research 114.
Filliben, James J. 1975. The Probability Plot Correlation Coefficient Test for Normality. Techno-
metrics 17 (1). https://doi.org/10.1080/00401706.1975.10489279.
Garrote, J., F.M. Alvarenga, and A. Díez-Herrero. 2016. Quantification of Flash Flood Economic
Risk Using Ultra-Detailed Stage–Damage Functions and 2-D Hydraulic Models. Journal of
Hydrology 541: 611–25. https://doi.org/10.1016/j.jhydrol.2016.02.006.
Haase, Dagmar. 2009. Effects of Urbanisation on the Water Balance—A Long-Term Trajectory.
Environmental Impact Assessment Review 29 (4): 211–219. https://doi.org/10.1016/j.eiar.2009.
01.002.
Hasan Tarek, Mehedi, Shah Md, Imran Kabir, and Adil Hassan. 2017. Application of Swmm for
Analysis of Flash Floods in Urban Areas: A Case Study for Chaktai Khal Watershed in Chittagong.
Hossain, M.M., and Rumman Mowla Chowdhury. 2018. Hydro-Morphological Study for Rehabil-
itation of Old Madhumati River Using Mathematical Model. Journal of Engineering Science 03
(1).
Hossain, Sharif, Guna Alankarage Hewa, and Subhashini Wella-Hewage. 2019. A Comparison
of Continuous and Event-Based Rainfall-Runoff (RR) Modelling Using EPA-SWMM. Water
(Switzerland) 11 (3). https://doi.org/10.3390/w11030611.
Institute of Water Modeling (IWM). 2011. Hydrological and Morphological Study of Old Madhu-
mati River under Rehabilitation of Old Madhumati River and Improvement of Surrounding Area
of Gopalganj Pourashava. Final Report (April). Dhaka.
IPCC. 2014. 5th Assessment Report: Summary for Policymakers. IPCC Fifth Asessement Report:
Working Group III Mitigation of Climate Change.
Jang, Suhyung, Minock Cho, Jaeyoung Yoon, Yongnam Yoon, Sangdan Kim, Geonha Kim,
Leehyung Kim, and Hafzullah Aksoy. 2007. Using SWMM as a Tool for Hydrologic Impact
Assessment. Desalination 212 (1–3): 344–356. https://doi.org/10.1016/j.desal.2007.05.005.
Khadka, Sanjay, and Keshav Basnet. 2019. Storm Water Management of Barahi Chowk Area,
Lakeside, Pokhara, Nepal Using SWMM. KEC Conference 2 (1): 320–325.
Khan, M.S.A., and J.U. Chowdhury. 1998. Dhaka City Storm Water Quality Assessment. Technical
Report-1. IWFM, BUET, Dhaka.
Khan, M.S., Md. Rezaur Rahman, Anisul Haque, Md. Munsur Rahman and G.M. Tarekul Islam.
2006. Stormwater Flooding at BUET Campus: A Situation Report. Institute of Water and Flood
Management, BUET, Dhaka.
Mohammed, Khaled, A.K.M. Saiful Islam, G.M. Tarekul Islam, Md., Lorenzo Alfieri, Jamal Uddin
Khan, Sujit Kumar Bala, and Mohan Kumar Das. 2018. Future Floods in Bangladesh under 1.5°C,
2°C, and 4°C Global Warming Scenarios. Journal of Hydrologic Engineering 23 (12): 04018050.
https://doi.org/10.1061/(asce)he.1943-5584.0001705.
Mohammed, Khaled, A.K.M. Saiful Islam, G.M. Tarekul Islam, Lorenzo Alfieri, Sujit Kumar Bala,
and Md. Jamal Uddin Khan. 2017. Impact of High-End Climate Change on Floods and Low
Flows of the Brahmaputra River. Journal of Hydrologic Engineering 22 (10): 04017041. https://
doi.org/10.1061/(asce)he.1943-5584.0001567.
Mondal, M. Shahjahan, A.K.M. Saiful Islam, Anisul Haque, Rashedul Islam, Subir Biswas, and
Khaled Mohammed. 2018. Assessing High-End Climate Change Impacts on Floods in Major
Rivers of Bangladesh Using Multi-Model Simulations. Global Science and Technology Journal
6 (2): 1–14.
Paul, S., A.K.M. Saiful Islam, G.M. Tarekul Islam, Ahsan Azhar Shopan, and Sujit Kumar Bala.
2013. Impact of Climate Change on Urban Drainage Systems in Three Selected Coastal Towns
of Bangladesh. International Conference on Climate Change Impacts and Adaptation 397–406.
Rabori, Ali Moafi, and Reza Ghazavi. 2018. Urban Flood Estimation and Evaluation of the
Performance of an Urban Drainage System in a Semi-Arid Urban Area Using SWMM.
Water Environment Research 90 (12): 2075–2082. https://doi.org/10.2175/106143017x15131012
188213.
Rahman, M.R., J.U. Chowdhury, and M. Salehin. 1999. Dhaka City Storm Water Quality
Assessment. Technical Report-2, R01/99. IWFM, BUET, Dhaka.
Rossman, Lewis A., and Wayne C. Huber. 2016. Storm Water Management Model Reference
Manual. US EPA Office of Research and Development III (January): 231. www2.epa.gov/water-
research.
Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor, and H.L. Miller.
2007. IPCC, 2007. Climate Change 2007: The Physical Science Basis. Contribution of Working
Group I to the Fourth Assessment Report of the Intergovermental Panel on Climate Change.
Flood Propagation Processes
in the Jamuna River Floodplain
in Sirajganj
Ashik Iqbal , M. Shahjahan Mondal, M. Shah Alam Khan,

Hans Hakvoort, and William Veerbeek
Abstract Monsoon flooding inundates a substantial part of Bangladesh, where 80%

of the areas are floodplains. Sirajganj, located beside the Jamuna River in north-
western Bangladesh, is home to many communities living in the low-lying unpro-
tected floodplains. The Ranigram village of Sirajganj has a hydraulic connection with
the Jamuna River and is flooded almost every year. This study aims to explore how
flood propagates in the floodplain, determine the water level variation in the flood-
plain with respect to the Jamuna River stage, and develop a hydrodynamic model
of the flood propagation process in Ranigram. Water level gauges were installed at
strategically selected locations in Ranigram, and flood data were collected during the
2018 and 2020 monsoons. In the R2 , NSE, MSE, and RMSE analyses, the observed
floodwater level in Ranigram shows an excellent dynamic relation with the water level
at Sirajganj on the Jamuna River. The statistical relation derived from the 2018 data
is validated with the observed data of 2020. A two-dimensional hydrodynamic model
is developed with HEC-RAS using a DEM generated with surveyed bathymetry and
UAV-based data and is calibrated and validated with the observed water level data
and flood images. The maps of flood arrival time, duration, and maximum extent and
depth were extracted from the model. The outcomes of this study will be useful in
flood risk and damage assessments and forecasting of floods in floodplain areas.
A. Iqbal (B) · M. S. Mondal · M. S. A. Khan

Institute of Water and Flood Management, Bangladesh University of Engineering and
Technology, Dhaka, Bangladesh
M. S. Mondal
e-mail: [email protected]
M. S. A. Khan
H. Hakvoort · W. Veerbeek
IHE Delft Institute for Water Education, Delft, The Netherlands
W. Veerbeek

46 A. Iqbal et al.
Keywords Flood propagation · Floodplain · Jamuna River · 2D hydrodynamic

model · HEC-RAS
1 Introduction
Almost 80% of Bangladesh is consisted of floodplain land, which is considered flood-

prone, and in average flood years, about 20% of the country’s land area is flooded
(Mirza 2002). Monsoon flooding from the river inundates a substantial part, espe-
cially the northern parts of Bangladesh, every year from early July to late September
(Khan 2008). Many communities reside in the low topographical non-protected area
of the riverside that is flooded by the river overtopping water. Such communities
are entirely exposed and vulnerable even to flood events with low associated return
periods. During floods, they are the first to be affected and cannot take enough precau-
tions before the floodwater enters the locality (Mirza 2002). Residents of these areas
often build their houses on elevated mounds as an indigenous technique to save their
houses from flood exposure (Brammer 2010).
Due to climate change and riverbed siltation, flood events are becoming increas-
ingly unpredictable and severe every year, and so the historical practice of living
with water is being threatened by rising flood impacts (Biswas 2008; Mondal et al.
2018). In addition, the encroachment of wetlands and floodplains and the inadequate
drainage capacity may increase the flood process’s severity at the local level (Islam
et al. 2010; Miller and Hutchins 2017; Avinash 2014). Bangladesh faces many chal-
lenges in flood management (Ali et al. 2018), where the implementation gap and poor
planning and design of flood mitigation measures make the management system inad-
equate and ineffective (Vaz 2020; Brammer 2010). Nevertheless, a proper assessment
of flooding is necessary for selecting and taking effective measures.
One of the most efficient non-structural flood impact mitigation measures is flood
forecasting and early warning systems (Subramanya 2008). Proper assessment of
the flooding process in inundated areas is a prerequisite for proper flood forecasting.
The Flood Forecasting and Warning Center (FFWC) of the Bangladesh Water Devel-
opment Board (BWDB) provides information of the river stage at critical locations
along the major rivers in Bangladesh (FFWC 2018). The FFWC has developed danger
levels for different locations to indicate the magnitudes of the expected flood condi-
tions (FFWC 2018). These levels primarily indicate major events when evacuation
and temporary relocation might be needed.
Sirajganj Sadar upazila (sub-district) has hydraulic interaction with the Jamuna
River and is flooded every year due to the overtopping of the river (Islam et al.
2010). Many riverside areas and communities are not considered for protection from
flooding with the embankment, and they suffer from the loss of houses and livelihoods
(Ali et al. 2018). People of the locality do not get any proper location-specific flood
information with indicative lead time. The flood information provided by FFWC
is only based on the river stage, but there is no flood information for the locality.
Nevertheless, better insight into the flood extent and depth distribution in the area is
Flood Propagation Processes in the Jamuna River … 47
essential for safety that impacts livelihoods since insight into local flood conditions is
essential for appropriate responses (Hassan and Shah 2008). If the flood propagation
process is appropriately understood, the consequent hazards for the flood in the
locality can be forecasted; thus, the necessary measures can be taken.
Two-dimensional hydrodynamic modeling is an effective method for gaining
insight into the propagation of floods (Meire et al. 2010). Such models typically
rely on remote sensing data with subsequent coarse digital elevation models for river
flooding (Rahman2015; Jain et al. 2018). Such practices are adequate for predicting
the flood extent for relatively large areas (Jain et al. 2018; Yalcin 2020) but often lack
the level-of-detail needed for sub-critical floods at the scale of local communities.
The main objective of this study is to explore the flood propagation process in the
floodplain in the Ranigram area of Sirajganj. The specific objectives are (a) to find
out the flood process in the Jamuna River floodplain through a statistical relationship
with the river stage and (b) to develop a 2D hydrodynamic model in HEC-RAS with
a UAV-enhanced high-resolution DTM, and to validate with field data at the local
level.
2 Floods in Bangladesh
Floods are Bangladesh’s most common natural disaster that affects people’s lives
and livelihoods. Bangladesh is located in the downstream of the three largest river
systems in the world—the Ganges, the Brahmaputra, and the Meghna. The flat
topography of the floodplain and the monsoonal climate make the country highly
vulnerable to floods. The principal source of floods in Bangladesh is the river water
from the overbank flows of these major river systems, which result from the runoff
of prolonged high-intensity rainfalls during the monsoon over Nepal, India, and
Bangladesh (Rahman and Salehin 2013). This massive discharge itself cannot be
drained to the Bay of Bengal due to the high outfall water levels (Rahman and
Salehin 2013). In addition, the construction of roads without sufficient cross drainage
openings, alignment of roads transverse to the natural drainage paths, blockage of
drainage channels by solid waste dumping, construction of cross-dams on some
rivers, inadequate capacity of drainage sluices, and loss of connectivity between
rivers and floodplains due to unplanned and unregulated anthropogenic activities are
amplifying the frequency and magnitude of flood events (Rahman and Salehin 2013).
Overflowing rivers annually flood about one-fifth to one-third of the country from
June to September. When the peak flows in the two major rivers, the Brahmaputra
and the Ganges, synchronize, extreme floods occur with extensive coverage and
duration (Rahman et al. 2006). An increasing variability is observed in the annually
flooded area since the late 1980s (Fig. 1). As seen in the figure, some severe floods
were experienced in 1988, 1998, 2007, and 2017, and some moderate floods in 1955,
1974, 1987, and 2004.
48 A. Iqbal et al.
Percentage of Flood Affected Area 80
60
40
20
0
1954
1957
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
2005
2008
2011
2014
2017
2020
Year
Fig. 1 Percentage of flood affected area in Bangladesh
3 Study Area
3.1 Description of the Area
Sirajganj is in the northwest of Bangladesh and is also one of the most flood-prone
areas. It lies between 24°22 and 24°37 N latitudes and between 89°36 and 89°47 E
longitudes. Out of 320 km2 of total area of Sirajganj region, about 260 km2 is desig-
nated as the land area and 60 km2 as the riverine area (Islam and Miah 2012). Siraj-
ganj is located along the Jamuna River and is flooded primarily by the riverine flood
during the monsoon associated with high-intensity rainfall. Sirajganj experiences
a sub-tropical monsoonal climate, typical of Bangladesh, with an average annual
rainfall of 1610 mm (BBS 2016). The danger flood level at the Sirajganj station on
the Jamuna River is set at 13.35 m PWD, which specifies that if this water level is
crossed, the resulting flood will cause damage in the adjacent areas.
The Ranigram village in the Khokshabari union is a peri-urban area of the Sirajganj
district and is adjacent to the Jamuna River (Fig. 2). The area of the village is 2.4 km2
and is located outside the new 15.35 m PWD flood-control embankment of the
Sirajganj town. It is surrounded on three sides by roads. Two crossbars exist in the
north and south of the village.
3.2 Flood Occurring in the Ranigram Area
There exists a breach in the eastern shallow old embankment (Fig. 3) from which
water enters the area. The Union Parishad water level gauge is just beside the water
Fig. 2 Study area with gauge locations and flooding direction
Fig. 3 Breaching location beside Union Parishad from where water enters into the Ranigram area
during a dry season and b monsoon
entry location. The breach profile is 10–30 m at the lowest portion of the breach and
extends to 110–170 m at the high flood level.
At first, the bank spilled water from the Jamuna River flows from the north to the
southeast side of the area between two crossbars. After that, water enters through the
breach, and for high flood levels, the overtopping occurs in the shallow embankment.
Then passing the breaching location, the water flows inside the village from south to
north and east to west from the Union Parishad. Gradually, water reaches the whole
northern part of the village and the culvert area in the middle-western part. From the
50 A. Iqbal et al.
culvert area, water flows to the southern part of the village and reaches the Pilot Site
area.
After entering the village, the water gets stacked because of the embankment
roads on all three sides of the village. With the decrease of water level in the Jamuna
River, the floodwater in the Ranigram village starts going out from the breaching
point location. The whole flood water goes out in the same way as it entered, and
little water remains in the ditches and lowlands that later dry up by evaporation.
4 Methodology
4.1 Methodological Framework
In Ranigram, three water level gauges are installed at strategically crucial locations to
monitor the floodplain’s water depth. From the flood depth, the water levels are calcu-
lated with proper benchmarking. The floodwater level data of the area is analyzed
to understand the relation of the internal flood process with the external river stage
of the Jamuna River by the coefficient of determination (R2 ), Nash–Sutcliffe model
efficiency coefficient (NSE), Mean Absolute Error (MAE), and Root Mean Square
Error (RMSE) analysis (Mali et al. 2020). Secondary data on the water level of the
Jamuna River is collected from BWDB.
A two-dimensional hydrodynamic model of the selected area is developed with
HEC-RAS using a high-resolution DTM (Farooq et al. 2019; Jung et al. 2014). Water
level and normal depth are used for boundary conditions for the model. Measured
flood water level data of one water level gauge is used for the boundary condition and
others for calibrating and validating the model (Yalcin 2020). The model is calibrated
and validated using real flood scenarios. The flood maps of the flood extent and depth,
arrival time, and duration are extracted from the simulation result. Figure 4 shows
the methodological framework of the study.
4.2 Data Collection
The water level readings of the three installed gauges were closely monitored for the
whole flood cycle of the years 2018 and 2020. The three water level gauges installed
at the Union Parishad, Culvert, and Pilot Site are shown in Fig. 2. The distance from
the Union Parishad gauge at the water entry location to the gauge at the Culvert is
about 0.82 km. The distance from the gauge at the Culvert to the Pilot site gauge is
about 1.35 km. The reduced level of each gauge location is measured with proper
benchmarking. When the water started overtopping the river, the water depth at each
gauge was collected regularly in the morning every day. Then, the water levels at
the three locations were measured by adding the depths with the respective reduced
(Boundary Condi ons)
Install Water Level Water Level and Normal Depth Surveyed

Gauges in the Bathymetry and
Floodplain Unmanned
2-D Hydrodynamic Aerial Vehicle -
Modelling with HEC-RAS Based DEM
Genera on
Collect Floodwater Calibra on and

Levels and Photos Valida on of Model
Daily
Compare Real- me Flood Photos at Selected
Water
Sta s cal Rela on Loca ons with the Model Flood Maps
Level of
with the Jamuna
the
River Stage
Jamuna
River Extract Flood Maps for Arrival Time, Depth and
Dura on of the Floods and Also Flood Depth and
Extent Maps of Different Return Period Floods
Fig. 4 Methodological framework of the study
levels. At the same time, the water level at Sirajganj station on the Jamuna River was
recorded daily from the FFWC website of BWDB.
4.3 Statistical Relation
A relationship between the water level on the floodplain and that on the river was
established using regression analysis. The simple linear regression method is used
for establishing the relationship between the water level on the floodplain and that on
the Jamuna River. The simple linear model is expressed using the following equation:
Y = aX + b + c (1)
where Y is the dependent variable (the water level at the gauges in the floodplain),
X is the independent variable (the water level on the Jamuna River at Sirajganj
station), a is the slope of the regression line, b is the intercept, and c is the residual
(error).
The R2 is the quantification of the predictive extent of a regression model. It
explains the strength of the relationship between the independent and dependent
variables. If the R2 of a model is 0.50, half of the observed variations are explained
52 A. Iqbal et al.
by the model’s inputs. The R2 values range from 0 to 1 and are commonly stated as
percentages from 0 to 100%. A higher R2 value indicates a better fit of the model to
the observed data (Freund et al. 2006).
4.4 Nash–Sutcliffe Model Efficiency Coefficient
The Nash–Sutcliffe model efficiency coefficient (NSE) is used to assess the predictive
skill of a model. It represents the robustness of the model. NSE value close to 1 means
the model quality is good, and if the value is negative, the model is unacceptable
(Zeybek 2018). The NSE is calculated as one minus the ratio of the error variance
of the modeled time-series divided by the variance of the observed time-series data.
It is calculated using the following equation:
n
(Yobs − Ysim )2
NSE = 1 − ni=1 (2)
i=1 (Yobs − Ymean )
2
where Y obs is the observed data, Y sim is the model data, and Y mean is the mean of
the observed data.
4.5 Mean Absolute Error
Mean Absolute Error (MAE) is an error function used for regression models. MAE is
the sum of absolute differences between the measured and predicted values. So, MAE
measures the average magnitude of errors in a set of predictions without considering
their directions. The formula is:
n
|Ysim − Yobs |
MAE = i=1 (3)
n
where n is the number of data, Y sim is the model data, Y obs is the measured data,
and |Y sim − Y obs | is the absolute error.
4.6 Root Mean Square Error
Root Mean Square Error (RMSE) is the standard deviation of the residuals or simply
called prediction errors. RMSE is a measure of how spread out these residuals are.
Thus, RMSE represents how robust the data is around the line of best fit.

n
(Ysim − Yobs )2
RMSE = i=1
(4)
n
4.7 High-Resolution DEM
A high-resolution DEM was generated using a hybrid technique with Unmanned

Aerial Vehicle (UAV) and measured bathymetry data of the Ranigram area. The
Agisoft Metashape 1.5.2 software and ArcGIS 10.5 software were used for the DTM
generation. The whole area was divided into two parts: (i) tree-canopy area and (ii)
non-canopy area. The UAV data was used for the non-canopy areas of the village.
The measured bathymetry data was used for the canopy areas which are covered by
dense trees where UAV cannot penetrate to take land position data.
For the non-canopy area, the UAV-generated images were used along with a set
of ground control points in the Agisoft Metashape 1.5.2 software to obtain a DTM
for the Ranigram area. Many errors were found when compared with the actual field
Fig. 5 High-resolution
DEM of Ranigram
54 A. Iqbal et al.
elevations. Some errors in the water-covered areas were corrected, and a manual
correction was done in the location of a hydraulic structure (i.e., culvert). For the
tree-canopy area, the measured bathymetry data was used. With this measured point
elevation data for the tree-canopy area, a complementary DTM was generated using
the Triangulated Irregular Network (TIN) methods from the point cloud. The two
DTM models of the tree-canopy and the non-canopy area were merged into one to
obtain the DTM of the Ranigram area (Fig. 5).
4.8 HEC-RAS 2D Model
The 2D hydrodynamic model was developed with HEC-RAS 5.0.7 software (Farooq
et al. 2019; Jung et al. 2014). The high-resolution DEM of the Ranigram area was
used in creating the terrain in the RAS Mapper. The projection was taken as WGS
1984—UTM zone 45 N. With satellite images, the horizontal position correctness
of the terrain is checked. At first, the 2D flow perimeter of the study area was set.
The shapefile of the Ranigram area was imported for the 2D flow area perimeter.
The mesh was developed with computational points in 10 m × 10 m for the 2D area.
Manning’s roughness coefficients of 0.04 and 0.05 were used for agricultural land
and housing land respectively in the 2D flow area.
Two external boundary condition lines were drawn. One boundary condition is
at the breaching location beside the Union Parishad from where the water enters
into the Ranigram area, and the other is in the south of the Ranigram village beside
the pilot site. For the first boundary condition, the stage hydrograph was taken, and
for the second boundary condition, the normal depth was taken. The daily flood
water level at the water entry location beside the Union Parishad was used for stage
hydrograph at boundary condition 1 with a 1-day interval. The average normal depth
for the second boundary condition estimated from the measured water level data at
the gauges and found to be 0.0001 was used for the floodplain of Sirajganj.
The model run was performed for the dates similar to the measured data. The
model was run with a 1-min computation interval. The output of the model was
taken for a 1-day interval. The water level at the Union Parishad was used as a
boundary condition, and the water level reading at the Culvert and Pilot Site was
used for calibrating and validating the model. The measured data of 2020 was used
for the validation of the model.
Fig. 6 Water level at Water Level Danger Level
Waterlevel Jamuna, Sirajganj(m PWD)

Sirajganj station in the 15.5
Jamuna River for different 15
return periods
14.5
14
13.5
13
12.5
12
1 10 100
Return period (years)
5 Results and Discussion
5.1 Frequency Analysis of the Jamuna River Water Level

at Sirajganj
Water level data at Sirajganj station of the Jamuna from 1984 to 2020 (37 years) has
been considered, and the Gumbel distribution is used for the frequency analysis. The
water levels for different return period floods are shown in Fig. 6. The danger level
is 13.35 m PWD. The Jamuna River crossed the danger level every year except for
five years in the 37 years observed.
For a 1.18-year return period, the water level in the Jamuna River at Sirajganj
crosses the danger level, and for a 2-year return period, the water level at the Jamuna
River is 0.65 m above the danger level.
5.2 Statistical Analysis of the Jamuna River Water Level

with the Flood Level in Floodplain
The flood depth data at each water level gauge in Ranigram village was collected
during the flood season of 2018. The floodwater level for the corresponding depth
was calculated for each day at each gauge. The water level at the Ranigram area
is plotted with the water level of the Jamuna at Sirajganj station in Fig. 7. The co-
relation with each gauge was found from the flood water level at the gauges in the
Ranigram area. It is found that when there was a sudden rise or fall in the water level
of the Jamuna River, the water level in the floodplain did not follow that trend.
56 A. Iqbal et al.
14
13.5
Water Level (m PWD)
13
12.5
12
11.5
11
Date
Jamuna(Sirajganj) Union Parishad Culvert Pilot site
Fig. 7 Flood water level data in the Ranigram and Sirajganj station of Jamuna River in 2018
5.2.1 The Relation of Gauges with the Water Level of the Jamuna
at Sirajganj
The water level data at the Ranigram area for each gauge was plotted with the water
level of the Jamuna at Sirajganj station. From the plots in Fig. 8a–c, it is found that
there is a linear relationship between the water level on the floodplain and the water
level on the Jamuna River, and so with the data, a statistical relationship can be
established using a simple regression method.
The statistical relation found between the Jamuna River water level and the water
level at the Union Parishad according to (1) is given below:
YUP = 1.1093X − 0.9437 (5)
YCulvert = 1.0635X − 0.3715 (6)
YPilotSite = 1.1072X − 1.0313 (7)
where Y is the water level at Union Parishad, Culvert, and Pilot Site in Eqs. (5),
(6), and (7), respectively, and X is the water level on the Jamuna at Sirajganj for
the corresponding day. The relation and error in the model were analyzed using the
Eqs. (2), (3), and (4).
The R2 value is 0.79, 0.83, and 0.86, and the NSE is 0.795, 0.831, and 0.864 for
Union Parishad, Culvert, and Pilot Site, respectively. In the error analysis, the MAE
Fig. 8 Relation of the 14.5

(a)
Union Parishad WL(m PWD)

Jamuna River water level y = 1.1093x - 0.9437
14
with the water level on the R² = 0.7957
floodplain at a Union 13.5
Parishad, b Culvert, and c
13
Pilot Site for 2018
12.5
12
11.5
11 11.5 12 12.5 13 13.5 14
Jamuna (Sirajganj) WL(m PWD)
14
y = 1.0635x - 0.3715
(b)
Culvert WL(mPWD)
13.5 R² = 0.831
13
12.5
12
11.5
11 11.5 12 12.5 13 13.5 14
Jamuna (Sirajganj) WL(m PWD)
14
y = 1.1072x - 1.0313
(c)
Pilot Site WL(m PWD)
13.5 R² = 0.8647
13
12.5
12
11.5
11
11 11.5 12 12.5 13 13.5 14
Jamuna (Sirajganj) WL(m pWD)
is found to be 0.017 m, 0.048 m, and 0.049 m, and the RMSE is 0.236 m, 0.201 m,
and 0.184 m for Union Parishad, Culvert, and Pilot Site, respectively. The residual
ranges from 0.267 m to −0.298 m, 0.276 m to −0.176 m, and 0.334 m to −0.240 m
for Union Parishad, Culvert, and Pilot Site, respectively.
The simple linear equation derived from the measured daily data of 2018 indicates
a good relation, and the model’s error is not very high. This model can predict water
levels at those locations from the water level of the Jamuna River. For low water
levels, the relationship deviates more, and for high water levels, the relationship is
better.
58 A. Iqbal et al.
5.2.2 Validation of Statistical Relation with 2020 Data
Equations (5), (6), and (7) developed with the data of 2018 were validated using
the measured data of 2020. First, the daily data of the Jamuna River of 2020 was
inputted as the independent variable for each equation, and from the equations, the
water level for each gauge was found. Figure 9a–c show the plot of the model and
measured water levels at Union Parishad, Culvert, and Pilot Site, respectively.
The R2 values are 0.975, 0.941, and 0.936, and the NSEs are 0.97, 0.92, and 0.87
for Union Parishad, Culvert, and Pilot Site, respectively.
Also, in the error analysis, the MAEs are found to be 0.137 m, 0.162 m, and
0.183 m, and the RMSEs are found to be 0.174 m, 0.267 m, and 0.301 m for Union
Parishad, Culvert, and Pilot Site, respectively. The residuals range from 0.418 m to
−0.392 m, 1.228 m to −0.370 m, and 1.355 m to −0.406 m for Union Parishad,
Culvert, and Pilot Site, respectively.
From Fig. 9, it is found that the statistical equations perform very well for the
2020 data. The variation of model and measured water levels at all three gauges is
very well captured, and the residuals are not very high.
5.3 HEC-RAS 2D Model Results
The two-dimensional HEC-RAS model incorporated with high-resolution DTM was

developed, and the simulation results were calibrated and validated with the measured
data of 2018 and 2020, respectively. The water level at the Union Parishad was used
as a boundary condition, and the water levels at the Culvert and Pilot Site were used
for calibrating and validating the model, respectively.
5.3.1 HEC-RAS Model Calibration
After the simulation, the computational cells at the Culvert and Pilot Site water level
gauge locations were selected. The daily water surface elevation data of the cells was
collected from the model output and is plotted with the measured water level of the
gauges. For the set up, the model performs well for the floodplain in Sirajganj.
Figure 10a, b show the plot of HEC-RAS simulated and measured water levels
at Culvert and Pilot Site, respectively. From the figure, it is seen that the simulated
water level is matching with the pattern of the measured water level at both Culvert
and Pilot Site. The R2 values are 0.982 and 0.975, and the NSEs are 0.97 and 0.91
for the Culvert and Pilot Site, respectively. The MAEs are found to be 0.043 m and
0.131 m, and the RMSEs to be 0.068 m and 0.141 m for the Culvert and Pilot Site,
respectively. The residuals range from 0.34 m to −0.07 m and 0.167 m to −0.271 m
for the Culvert and Pilot Site, respectively.
So, from the calibration, it appears that the HEC-RAS model incorporated with
high-resolution DTM performs very well when compared with the measured water
16
(a)
Water Level (m PWD)

15
14
13
12
11
10
Date
14-Oct-2020
27-May-2020
6-Jul-2020
26-Jul-2020
15-Aug-2020
16-Jun-2020
4-Sep-2020
24-Sep-2020
UP_2020 from Regression UP Measured _ 2020
16
(b)
Water Level (m PWD)
15
14
13
12
11
10
Date
Culvert 2020 from Regression Culvert Measured _ 2020
16
(c)
Water Level (m PWD)
15
14
13
12
11
10
Date
Pilot site 2020 from Regression Pilot site Measured _ 2020
Fig. 9 Model water level with the measured water level in the floodplain at a Union Parishad, b
Culvert, and c Pilot Site for 2020
60 A. Iqbal et al.
14.5
(a)
14
Water Level (m PWD) 13.5
13
12.5
12
11.5
11
10.5
Date
Measured WL at Culvert _2018 WL from HEC-RAS Simula on; n:0.05
14.5
14 (b)
Water Level (m PWD)
13.5
13
12.5
12
11.5
11
10.5
Date
Measured WL at Pilot site _2018 WL from HEC-RAS Simula on; n:0.05
Fig. 10 HEC-RAS model calibration with simulated water level and measured water level in
floodplain at a Culvert and b Pilot Site for 2018
level in the Ranigram area. Though there is some residual in the model, it is not very
high.
5.3.2 HEC-RAS Model Validation
The water level data for 2020 at the Union Parishad derived from Eq. (5) was used
in the HEC-RAS model. The 2D HEC-RAS model was run with the water level of
2020 as the boundary condition at the Union Parishad location. Figure 11a, b show
the plots of HEC-RAS simulated and measured water levels in 2020 at the Culvert
and Pilot Site, respectively.
The R2 values are 0.963 and 0.948, and the NSEs are 0.82 and 0.87 for the Culvert
and Pilot Site, respectively. The MAEs are 0.055 m and 0.189 m, and the RMSEs
15.5
Water Level (m PWD)

(a)
14.5
13.5
12.5
11.5
10.5
14-Oct-2020
27-May-2020
6-Jul-2020
26-Jul-2020
15-Aug-2020
16-Jun-2020
4-Sep-2020
24-Sep-2020
Date
WL from HEC-RAS Simula on_2020 Measured WL at Pilot Site _2020
15.5
15 (b)
Water Level (m PWD)
14.5
14
13.5
13
12.5
12
11.5
11
10.5
14-Oct-2020
27-May-2020
6-Jul-2020
26-Jul-2020
15-Aug-2020
16-Jun-2020
4-Sep-2020
24-Sep-2020
Date
WL from HEC-RAS Simula on_2020 Measured WL at Culvert _2020
Fig. 11 HEC-RAS model validation with simulated water level and measured water level in
floodplain at a Culvert and b Pilot Site for 2020
are 0.12 m and 0. 0.227 m for the Culvert and Pilot Site, respectively. The residuals
range from 0.691 m to −0.08 m and 0.338 m to −0.667 m for the Culvert and Pilot
Site, respectively.
From the analysis, it is found that the model performs very well when validated
with the measured data of 2020. The residual is high at the beginning and end of the
flood period, which means that in those days there is a sudden rise or fall in the water
level in the Jamuna River.
62 A. Iqbal et al.
Fig. 12 Flood maps for 28 June 2020 with real-time photos in the floodplain of Ranigram
5.3.3 Model Comparison with Real-Time Photos of Some Specific

Locations of Ranigram
The result of the model was also compared with an actual flood event in 2020. The
photos were collected from crucial locations in Ranigram during the flood. On 28 June
and 15 July 2020, photos were taken from the local market beside the Union Parishad,
a retrofitted house, Pilot Site, and Culvert. The flood maps for the 28 June 2020 and
15 July 2020 were also exported from HEC-RAS, and the maps were compared at
the locations of the photos. These are shown in Figs. 12 and 13, respectively.
The model-simulated flood maps show a similar flood scenario when compared
with the photos of actual floods. Thus, the model with a high-resolution DEM can
represent the real flood scenario at the locality of the floodplain.
5.4 Maps of Different Flood Parameters
The maps of flood arrival time, duration, and maximum depth and extent for the
Ranigram area were extracted for both the 2018 and 2020 flood events (Fig. 14).
At first, flooding started in the low-elevated croplands and reached the high-elevated
housing within the following two to three weeks. The maximum duration of the flood
was three months and twenty days, and four months and twenty days for 2018 and
2020, respectively. The flood information for any specific location in the village can
be derived using these maps. The inundation depth in the wetlands is more than 9 m.
The agricultural lowlands inundated for about 3.5 m, and the housing lands inundated
Fig. 13 Flood maps for 15 July 2020 with real-time photos in the floodplain of Ranigram
Fig. 14 Flood maps of a and d arrival time, b and e duration, and c and f maximum depth and
extent for the flood events of 2018 and 2020, respectively
64 A. Iqbal et al.
for more than 1 m. The flood events in the Ranigram area were severe in the years
2018 and 2020.
5.5 Flood Maps of Ranigram for Different Return Periods
The flood maps for different return periods are generated for Ranigram using the
HEC-RAS model. Figure 15 shows the flood depth and extent maps of the Ranigram
area for 1-, 2-, 10-, 20-, 50-, and 100-year return periods. From the maps, it is found
that the agricultural lowlands get flooded every year. For a 2-year return period flood,
the model shows complete flooding of the Ranigram area. During a 10-year flood
event, the mounds with elevated houses are inundated for a depth of about 1.00–
1.25 m, while during a 50-year flood the inundation depth is about 1.25–1.50 m.
However, the inundation depth increases by 0.03 m only from a 50- to a 100-year
return period flood.
Fig. 15 Flood depth and extent maps for return periods of a 1 year, b 2 years, c 10 years, d 20 years,
e 50 years and f 100 years in the Ranigram area
6 Conclusion
The flood in Ranigram area is found to be influenced almost entirely by the waters of
the Jamuna River. Flooding in Ranigram increases with the corresponding increase in
the Jamuna water levels and vice versa. The relation of floodwater level in the riverside
area is linear and provides a good fit with the water level of the Jamuna River. The
statistical equation can be useful in predicting the water levels at the gauge locations
in the Ranigram area. A 2D hydrodynamic model with high-resolution DEM can
represent a more accurate flood scenario at the local level. More field data frequency
and UAV and survey bathymetry-based DEM can contribute to the accuracy level of
the model. The flood maps representing arrival time, duration, and maximum depth
and extent for the Ranigram area can be extracted from the hydrodynamic model. The
flood information for any specific location of the floodplain can be obtained from the
flood maps. This information can be helpful in flood risk and damage assessments
and forecasting of floods in floodplain areas.
Acknowledgements Alhamdulillahi Rabbil Alamin. We praise and glorify the Lord, Allah
Subhanahu wa taala, as He is ought to be praised and glorified. Authors would like to express grat-
itude to local NGO SHARP, Sirajganj for their cordial logistic support during field visit and data
collection. This research is a part of the project “Community Resilience through rapid prototyping
of flood proofing technologies in urban areas (CORE Bangladesh)” in the Urbanizing Deltas of
the World-2 Program supported by The Netherlands Organization for Scientific Research (NWO)
under project number W07.69.203.
References
Ali, Md.H., B. Bhattacharya, A.K.M.S. Islam, G.M.T. Islam, Md.S. Hossain, and A.S. Khan. 2018.
Challenges for Flood Risk Management in Flood-Prone Sirajganj Region of Bangladesh. Journal
of Flood Risk Management 12 (1): e12450. https://doi.org/10.1111/jfr3.12450.
Avinash, Shubha. 2014. Flood Related Disasters: Concerned to Urban Flooding in Bangalore, India.
International Journal of Research in Engineering and Technology 03 (28): 76–83. https://doi.org/
10.15623/ijret.2014.0328013.
BBS. 2016. Economic Census 2013—District Report: Sirajganj. Reproduction, Documentation &
Publication (RDP) Section Bangladesh Bureau of Statistics, Bangladesh.
Biswas, Asit K. 2008. Management of Ganges-Brahmaputra-Meghna System: Way Forward.
In Management of Transboundary Rivers and Lakes, 143–164. https://doi.org/10.1007/978-3-
540-74928-8_6.
Brammer, Hugh. 2010. After the Bangladesh Flood Action Plan: Looking to the Future.
Environmental Hazards 9 (1): 118–130. https://doi.org/10.3763/ehaz.2010.si01.
Farooq, Muhammad, Muhammad Shafique, and Muhammad Shahzad Khattak. 2019. Flood Hazard
Assessment and Mapping of River Swat Using HEC-RAS 2D Model and High-Resolution 12-M
Tandem-X DEM (Worlddem). Natural Hazards 97 (2): 477–492. https://doi.org/10.1007/s11069-
019-03638-9.
FFWC. 2018. Annual Flood Report 2018. Processing and Flood Forecasting Circle, Bangladesh
Water Development Board. Government of the People’s Republic of Bangladesh.
66 A. Iqbal et al.
Freund, Rudolf J., William J. Wilson, and Ping Sa. 2006. Regression Analysis, 2nd ed. Boston:
Academic Press.
Hassan, Ahmadul, and Mohammad Aminur Rahman Shah. 2008. A Participatory Model for Flood
Early Warning Dissemination. In Third South Asia Water Research Conference On Innovative
Modeling Approaches For IWRM. Dhaka.
Islam, Sirajul, and Sajahan Miah. 2012. Banglapedia: National Encyclopedia of Bangladesh. Dhaka:
Asiatic Society of Bangladesh.
Islam, Akm Saiful, Anisul Haque, and Sujit Kumar Bala. 2010. Hydrologic Characteristics of Floods
in Ganges–Brahmaputra–Meghna (GBM) Delta. Natural Hazards 54 (3): 797–811. https://doi.
org/10.1007/s11069-010-9504-y.
Jain, Sharad Kumar, Pankaj Mani, Sanjay K. Jain, Pavithra Prakash, Vijay P. Singh, Desiree
Tullos, Sanjay Kumar, S.P. Agarwal, and A.P. Dimri. 2018. A Brief Review of Flood Forecasting
Techniques and Their Applications. International Journal of River Basin Management 16 (3):
329–344. https://doi.org/10.1080/15715124.2017.1411920.
Jung, Younghun, Dongkyun Kim, Dongwook Kim, Munmo Kim, and Seung Lee. 2014. Simplified
Flood Inundation Mapping Based on Flood Elevation-Discharge Rating Curves Using Satellite
Images in Gauged Watersheds. Water 6 (5): 1280–1299. https://doi.org/10.3390/w6051280.
Khan, M. Shah Alam. 2008. Disaster Preparedness for Sustainable Development in
Bangladesh. Disaster Prevention and Management: An International Journal 17 (5): 662–671.
https://doi.org/10.1108/09653560810918667.
Mali, Vijay Kisan, B. Veeranna, Aditya Parik, and Soumendra Nath Kuiry. 2020. Experimental
and Numerical Study of Flood Dynamics in a River-Network-Floodplain Set-Up. Journal of
Hydroinformatics 22 (4): 793–814. https://doi.org/10.2166/hydro.2020.160.
Meire, Dieter, Liesbet De Doncker, Frederic Declercq, Kerst Buis, Peter Troch, and Ronny Verho-
even. 2010. Modelling River-Floodplain Interaction During Flood Propagation. Natural Hazards
55 (1): 111–121. https://doi.org/10.1007/s11069-010-9554-1.
Miller, James D., and Michael Hutchins. 2017. The Impacts of Urbanisation and Climate Change
on Urban Flooding and Urban Water Quality: A Review of the Evidence Concerning the United
Kingdom. Journal of Hydrology: Regional Studies 12: 345–362. https://doi.org/10.1016/j.ejrh.
2017.06.006.
Mirza, M. Monirul Qader. 2002. Global Warming and Changes in the Probability of Occurrence of
Floods in Bangladesh and Implications. Global Environmental Change 12 (2): 127–138. https://
doi.org/10.1016/s0959-3780(02)00002-x.
Mondal, M. Shahjahan, A.K.M. Saiful Islam, Anisul Haque, Md. Rashedul Islam, Subir Biswas,
and Khaled Mohammed. 2018. Assessing High-End Climate Change Impacts on Floods in Major
Rivers of Bangladesh Using Multi-Model Simulations. Global Science and Technology Journal 6
(2): 1–14. https://zantworldpress.com/wp-content/uploads/2018/10/1.Mondal.pdf.
Rahman, Md. Mostafizur. 2015. Modeling Flood Inundation of the Jamuna River. M. Sc.
thesis, Department of Water Resources Engineering, Bangladesh University of Engineering and
Technology.
Rahman, Rezaur, and Mashfiqus Salehin. 2013. Flood Risks and Reduction Approaches in
Bangladesh. In Disaster Risk Reduction, 65–90. https://doi.org/10.1007/978-4-431-54252-0_4.
Rahman, Rezaur, Anisul Haque, M. Shah Alam Khan, Mashfiqus Salehin, and Sujit Kumar Bala.
2006. Investigation of Hydrological Aspects of Flood-2004 with Special Emphasis on Dhaka City.
Final Report, Institute of Water and Flood Management, Bangladesh University of Engineering
and Technology, Dhaka.
Subramanya, K. 2008. Engineering Hydrology, 3rd ed. New Delhi: Tata McGraw-Hill.
Vaz, A. C. 2020. Coping with floods—the experience of Mozambique. 1st WARFSA/WaterNet
Symposium: Sustainable Use of Water Resources, 1–2 November 2000, Maputo.
Yalcin, Emrah. 2020. Assessing the Impact of Topography and Land Cover Data Resolutions
on Two-Dimensional HEC-RAS Hydrodynamic Model Simulations for Urban Flood Hazard
Analysis. Natural Hazards 101 (3): 995–1017. https://doi.org/10.1007/s11069-020-03906-z.
Zeybek, Melis. 2018. Nash-Sutcliffe Efficiency Approach for Quality Improvement. Journal of
Applied Mathematics and Computation 2 (11): 496–503. https://doi.org/10.26855/jamc.2018.
11.001.
Co-creation of Flood Mitigation
Technologies in Bangladesh
to Strengthen Community Resilience
Nadia Nowshin, M. Shah Alam Khan, Hans Hakvoort, William Veerbeek,

and Chris Zevenbergen
Abstract Most of the small- to medium-scale flood mitigation technologies in

Bangladesh are planned, designed, and constructed in a top-down approach, ignoring
the actual needs of the vulnerable communities and the appropriateness of the tech-
nology in the local context. This non-inclusive approach does not create owner-
ship of the technology by the local communities and stakeholders, and undermines
the sustainability of the technological solution. This paper presents the co-creation
process and preliminary performance monitoring results of a few small-scale flood
mitigation technologies implemented as pilots in the peri-urban area of Sirajganj,
Bangladesh. Local communities, policy actors, and stakeholders were involved at
various stages of planning, design, construction, management, and monitoring of
these pilots. Sustainability scores of the technologies in their technical, economic,
social, and environmental dimensions are assessed through a set of indicators. A
comparative sustainability analysis of these technologies is performed based on
the user experiences and expert opinions. It is concluded that successful technical
performance and functional effectiveness, along with reasonable and fair costs, are
the primary requirements for the sustainability of a flood mitigation technology to
improve community resilience. This study also provides key policy recommenda-
tions for building community resilience in urban/peri-urban areas of Bangladesh to
deal with flood risks.
Keywords Flood mitigation · Community resilience · Technology · Co-creation ·

Sustainable development
N. Nowshin (B) · M. S. A. Khan

H. Hakvoort · W. Veerbeek · C. Zevenbergen
IHE Delft Institute for Water Education, Delft, The Netherlands

70 N. Nowshin et al.
1 Introduction
Bangladesh is one of the largest floodplain and delta areas in the world and is highly
vulnerable to floods due to its geographical location at the confluence of the Ganges,
Brahmaputra and Meghna (GBM) rivers. The rapidly increasing population, the
intensification of agriculture and the scale and dynamics of the river systems make
floodplain management in Bangladesh a challenging task (Mirza 2003; Brammer
2004; Islam 2017). Although flooding is a natural process, human activities enhance
the risks of flooding. Human activities that usually cause flooding include land use
changes, clearance of vegetation, cultivation in steep slope areas, rapid and unplanned
urbanization, poor design and maintenance of drainage system and settlement in
floodplain areas (Sinthumule and Mudau 2019; Flower and Fortnam 2015; Kabir
and Hossen 2019).
About 80% of Bangladesh is occupied by floodplain and is affected by river floods,
rainwater floods and flash floods (Brammer 1990). Floodplains in Bangladesh are
subject to riverine floods during the monsoon while the urban and peri-urban areas
experience a combined effect of rainfall-induced floods and riverine floods. Flood
damages become the most extensive when the three major rivers reach their peak
stages at the same time. Rainfall-induced floods are caused by high-intensity local
precipitation of long duration during the monsoon. It is projected that increased
precipitation, higher transboundary water flows and sea level rise will increase the
destructive power of monsoon floods in future, mainly in embanked areas (Dasgupta
et al. 2011; Dewan et al. 2003). This will ultimately cause serious threats to the
lives and livelihoods, particularly for the poor floodplain inhabitants and low income
groups in urban and peri-urban areas (Sultana and Thompson 2019; Parvin et al.
2016).
Floods are the most widespread disaster risks to urban settlements of all sizes.
According to the UN Office for Disaster Risk Reduction (UNDRR 2020), 4.03 billion
people have been affected by floods over the last 20 years (2000–2019) due to the
combined effect of population growth and climate change-induced sea level rise,
storm surge, etc. The losses are expected to increase. To mitigate future losses,
building flood resilience at the community level is an important component of an
integrated flood risk management strategy that calls for measures that are customized
to the local needs and often inspired by indigenous technologies (Keating et al. 2014).
Sustainable development as articulated in the ambitious UN agenda for global
action on sustainable development calls for innovation. In developing countries, there
is a growing recognition of the importance of experimentation and demonstration of
local technological innovations at grassroots levels and marginalized communities,
and to involve them in the innovation process (Smith et al. 2016; Goodman et al.
2017). Empowering local communities through their participation in flood manage-
ment is an effective approach for successful flood governance practices over the
long term. Additionally, sustainability needs to be considered during improvement
Co-creation of Flood Mitigation Technologies in Bangladesh … 71
of flood resilience at local levels. Moreover, capitals should be utilized at commu-

nity level through introduction of new, innovative and improvised technology, and
enhancement of existing recovery practices of that community (Corsel 2015).
In Bangladesh, this particularly holds true for sustainable (or local) flood miti-
gation technologies where the communities also represent a significant market for
small-scale entrepreneurs. Community stakeholders can help in finding an effec-
tive and sustainable intervention that matches each community’s unique context and
demand. Success crucially depends on the communities’ involvement in the plan-
ning and decision-making processes, as well as on their engagement over the long
term. Identifying the right interventions for boosting community resilience is only
one part of the equation; other aspects include the local engagement in carrying them
out (Deubelli 2019). Within this context, innovations in flood mitigation technolo-
gies often comprise of a process of incremental steps based on adjusting to existing,
indigenous technologies, rather than on a radical new technology.
In general, the criteria for sustainable flood mitigation technology are: (i) technical
and functional effectiveness, i.e., the capacity of measures to fulfill the primary
function of risk management; (ii) economic sustainability, i.e., affordability; (iii)
social sustainability, i.e., social acceptability; (iv) environmental sustainability, which
assesses the impact of measures relative to all species, habitats and landscapes; and
(v) institutional sustainability, which addresses the issue of governance (Edjossan-
Sossou et al. 2014).
It follows from the above that the development of sustainable flood mitigation
technologies should follow an inclusive co-creation process to effectively reduce
flood vulnerability and to comply with long-term technical, economic and social
goals (Montz and Grunfest 2002).
In this paper, we assess the sustainability of different flood mitigation technolo-
gies: Amphibious living unit, amphibious sanitation unit and retrofitted housing, etc.
(for small- to medium-scale floods affecting local communities) by using a set of
sustainability indicators (i.e., technical, social, economic and environmental) through
a co-creation process in a demonstration site. The outcomes of the assessment have
been used to develop policy recommendations for building community resilience in
urban/peri-urban areas.
2 Study Area
Sirajganj district is one of the most vulnerable flood-prone areas of Bangladesh.

Flood and water logging are some of the major disasters occurring in this area.
The main cause of flooding in the area is the transboundary inflow from upstream
catchment carried by the Jamuna River. Most of the upazilas are exposed to floods and
subsequent river erosion. About 65% of the people of Sirajganj district are vulnerable
to flooding (Rouf 2015).
Fig. 1 Map showing the study area
The demonstration (pilot) site is located in Sirajganj (Fig. 1), one of the most
flood-prone urban areas of Bangladesh. Yet this area is representative of the peri-
urban regions in the floodplains of a large river, the Jamuna. Figure 2 shows the
highest observed water level from 1984 to 2020 (37 years) in the Jamuna River at
Sirajganj Station where the danger level is 13.35 m PWD. The figure shows that
Fig. 2 Graph showing highest water lever in the Jamuna River at Sirajganj (Source Flood
Forecasting and Warning Center, FFWC)
the water level crosses the danger level almost every year, except for the five years
during this time period.
The study area is flooded for 140 days every year on an average which severely
affects the local living condition and livelihoods (Fig. 3a). Since these areas face
flood for almost five months per year, conducting normal year-round activities by the
inhabitants of these floodplain areas is quite challenging. They face different types of
socio-economic problems particularly during flood. Scarcity of pure drinking water
has become acute in flood-affected areas as the tube-wells have been submerged.
Fig. 3 a Sirajganj Peri-urban area during flood, b amphibious living unit, c amphibious sanita-
tion unit, d retrofitted house, e focus group discussion with user and manufacturer group, f local
community participation during construction (Source CORE BD Team)
The inhabitants of the Sirajganj peri-urban area are involved in numerous low-
paid income-generating activities such as day-laborers, carpenters, shop keepers, to
name of a few. As they face flooding every year, they are trapped in a vicious cycle
of poverty and hunger. During floods, the affected people have to put in all their
efforts for managing food for survival. Due to the lack of savings, they can barely
manage other basic needs and are forced to take personal loans that lead to more
difficult conditions. Poor conditions during this period lead to various sicknesses,
and they need to spend money for treatment purposes as well. They need pure and
safe drinking water, ORS (oral rehydration solutions), sanitation facilities and non-
food items (like sari, lungi, napkin/towel, mosquito net and lights at dark) along with
emergency food support.
During the flood period, when water enters into the houses, the household belong-
ings are completely damaged. Following the flood, they need to repair houses, sani-
tation units and other household commodities on a large scale. Due to the lack of
savings, it is quite difficult for them to spend additional money for repairing their
household items. Despite facing poverty, they have to repay this loan, which is a
burden for them.
To study community resilience, specific households are selected from this area.
Taking into account the vulnerabilities of the assigned households, this study offers
three innovative small-scale flood mitigation technologies such as amphibious living
unit (Fig. 3b), amphibious sanitation unit (Fig. 3c), and retrofitted house (Fig. 3d)
in order to reduce the flood risks and enhance community flood resilience (CORE
Bangladesh Project 2016).
3 Materials and Methods
Focus group discussions (FGDs) were conducted with users, manufacturers and
flood-affected communities (Fig. 3e) to collect qualitative data and information.
They shared their practical experiences during the construction period of these tech-
nologies and two flood periods (2019–2020) considering the social, economic and
environmental governance and technical aspects. This qualitative information was
converted to quantitative data by scores given on the basis of community feedback
and the researcher’s own perception and analysis. Finally, a comparative analysis
was carried out to determine the performance and sustainability of the technolo-
gies. In addition, several meetings were carried out with different stakeholders,
government officials and workshop owners. Workshops were conducted to exchange
views between technology providers and the local community. Students from Siraj-
ganj Polytechnic Institute participated in a design competition on the flood-proofed
(amphibious) house. Moreover, the local community participated in the construction
of the pilot flood-proofed houses (Fig. 3f).
Table 1 Stakeholder description and role

Stakeholder Description Role
Manufacturer Dutch SMEs (flood mitigation Design and construction of the pilot
technology providers), local technology, provide construction
mechanics and workshops, local materials locally
construction material suppliers
Implementer Netherlands Organization for Financial, academic, research, technical
Scientific Research, IHE-Delft, and field support
Bangladesh University of
Engineering and Technology, Dutch
SMEs
User Flood-affected community Participated in planning, design and
construction; provided land for pilot
construction; provided local knowledge
on flood processes and feasible options
for mitigation
Policy maker Local Government Engineering Influence and implement flood
Department (LGED), Bangladesh mitigation policy at the local and
Water Development Board (BWDB), national levels; adoption of the
Mayor (City Corporation), Union technology for up-scaling
Parishad Chairman (UP Chairman),
Local Leaders, Govt. Officials, Ward
Commissioner, Socio-Economist and
Technologist
3.1 Stakeholder Engagement
Stakeholders who had a role or interest in flood mitigation technology and community
resilience were identified through extensive discussion, interviews and field visits.
Some of these stakeholders were engaged with the research team from the inception;
others were engaged at a later stage. The stakeholders were classified into four groups:
Manufacturer, Implementer, User and Policy Maker on the basis of their activities
and importance. Table 1 gives the descriptions and roles of these stakeholders.
3.2 Assessment of Resilience and Sustainability
Several flood mitigation technologies, i.e., amphibious living unit, amphibious sani-
tation unit, retrofitted house, etc., were introduced to test sustainability of flood
mitigation solutions and to strengthen community resilience. The fundamental basis
of this approach was to ensure co-creation of the technologies through a participa-
tory process that would create ownership of the technologies and thus contribute
to sustainability and resilience. Figure 4 shows the conceptual framework of this
approach based on preliminary understanding of different factors that influence
Introduction of different
flood mitigation
technology
Social, economic,
Sustainable flood Flood adaptation and technical and
resilient area or zone transformation capacity environmental
sustainability
Effective
governance and up-
scaling
Fig. 4 Conceptual framework for resilience and sustainability of the co-creation process
sustainability and resilience. Subsequent verification and improvement of this frame-

work and selection of indicators were carried out through a detailed discussion with
stakeholders.
The framework shows that the adaptive and transformative capacities of the flood-
affected communities are enhanced through the introduction of innovative and afford-
able technologies, effective governance and scaling up of these technologies. Scale up
of these technologies will be ensured by proper governance and appropriate participa-
tion of policy actors in the decision-making process. The social, economic, technical
and environmental sustainabilities of these technologies are important indicators that
ultimately create the sustainable flood resilient area or zone.
3.3 Sustainability Dimensions and Indicators
Several indicators are selected to assess the sustainability of the flood mitigation
technologies (Table 2). In this study, technical sustainability is considered as the
most important dimension. The basic concept used here is that if a technology does
not perform successfully, it will not sustain in the long run even if it is less expensive,
socially acceptable and environment friendly. The order of importance of the other
dimensions presented in Table 2 is determined with similar justifications.
Table 2 Sustainability indicators for different dimensions

Dimension Indicator
Technical Availability of local materials, availability of local labor, construction ability,
technical performance, functional effectiveness
Economic Return on investment, business agility, life-cycle costs, willingness to pay,
employment, total capital cost, construction time
Social Labor practices and decent work, society and customers, human health risk
impact, acceptability to stakeholders, participation and responsibility of
different stakeholders, wellbeing/happiness, quality of life, flexibility and
adaptation
Environmental Impact on environment, waste, material and resource (reusability and waste),
pollution (land and water), impact of land use in ecosystem
3.4 Data Analysis
Flood mitigation technologies are evaluated in three steps. In the first step, a scoring
scale is set for different sustainability indicators to evaluate the technologies. The
scoring scale ranges from +2 to −2 (very good, good, neutral, bad and very bad),
where positive and negative scores indicate favorable and adverse impacts of the
technology, respectively.
In the second step, relative weights of the sustainability indicators are assigned on
ad hoc basis (Canter 1996) based on public consultation and expert judgment. The
weighted average score (I) (Eq. 1) for each alternative technology is determined for
each domain (Technical, Economic, Social and Environmental) in this step.

(W ∗ Score)
I = (1)
W
where I = weighted average score, W = relative weight (Range: 1–3), and score
= value assigned to each indicators on the basis of technical, economic, social and
environmental performance of the technology based on information collected through
field visits, focus group discussion (FGDs) and interviews.
Finally, the combined impact scores for the four dimensions are calculated to
identify the best performing technology with respect to sustainability. In this step,
relative weights of different dimensions are assigned based on unranked pair-wise
comparison technique. Pair-wise comparison technique (unranked and ranked) for
importance weighting basically involves a series of comparisons between decision
factors or criteria. This technique allows the comparison of two criteria or factors at a
time (Sims et al. 1991; Canter 1996; Elsheikh et al. 2015). In this study, this method
compares two alternative choices which are expected to have independent impacts
by ranking the parameters in pairs. Different combinations of the parameters are thus
tested to find out the most influential parameter(s) and thus deciding on the relative
weights (Canter 1996).
4.1 Evaluation of Alternative Technologies
Performance of these selected technologies are evaluated for the four dimensions,
i.e., technical, economic, social and environmental.
The first technology is an amphibious living unit (Fig. 5a, b) which is built on
an amphibious floating base where a special type of construction material (EPS—
Expanded Polystyrene) is used. The house floats during flood and sits on a level
ground after the flood water recedes.
The second technology is an amphibious sanitation unit (Fig. 5c, d). The floating
base and technology in the sanitation unit are the same as in the amphibious living
unit. This unit consists of two parts. The upper part is the toilet unit, while the
lower part contains a constructed wetland (which filters the discharged wastewater
by bioremediation) and a septic tank.
The third technology is a retrofitted house (Fig. 5e, f) in which a vertically-moving
wooden floor is constructed as a floating base. PET (polyethylene terephthalate)
bottles are used as the floating material in the wooden floor which moves up and
down with the floodwater level.
Performance of these technologies are assessed through field visits, FGDs and
interviews with the users and manufacturers. Scores and relative weights of the
indicators are assigned in the four dimensions for each technology based on these
interactions and expert judgment. Table 3 gives the weighted average scores of the
technical indicators of sustainability for the amphibious living unit, amphibious sani-
tation unit and retrofitted house. For amphibious living unit, considering different
advantages and disadvantages, the weighted average score is +0.6. Although the
most important material, i.e., EPS is not locally available and is expensive, tech-
nically, however, the EPS base performed very well. The base floated immediately
with the flood and perfectly set on the ground after the floodwater drained out. Since
no anchor was in place initially, the EPS base needed some manual maneuvering
to settle it down on the ground, although minor rolling was experiences. The local
community and the manufacturers can build the superstructure easily. However, they
need assistance to acquire the EPS and design the base. Moreover, EPS is found to
degrade with prolonged submergence in water and rodent attack. Local manufac-
turers are able to repair the wooden frame, but are unable to replace the EPS blocks
inside the frame. The superstructure remained stable and performed satisfactorily
during the flood event. The users report a very high load-bearing capacity for this
technology.
The weighted average score for the amphibious sanitation unit is −0.2. This score
reflects the same difficulties with EPS as in the living unit, but also the additional
difficulties with repair and maintenance. The superstructure and the appliances get
damaged more rapidly, sometimes making it unusable. The local mechanics and
plumbers are unable to fix these problems. The wooden superstructure is not suitable
for the local weather and gets damaged rapidly.
Fig. 5 a Schematic plan of amphibious living unit, b schematic elevation of amphibious living unit,
c schematic plan of amphibious toilet unit with constructed wetland, d schematic section elevation
of amphibious toilet unit with constructed wetland, e plan below the retrofitted wooden floor, f
elevation of section A-A of retrofitted floor (Note All designs are not to scale)
80
Table 3 Weighted average scores for the technical indicators of sustainability

Amphibious living unit Amphibious sanitation unit Retrofitted house
Indicators Relative weight (W) Score W * Score Relative weight (W) Score W * Score Relative weight (W) Score W * Score
Availability of 1 −1.25 −1.25 1 −1.5 −1.5 1 2 2
local materials
Availability of 1 1.25 1.25 1 −0.25 −0.25 1 2 2
local labor
Construction 1 −1.25 −1.25 1 −1.5 −1.5 1 1.25 1.25
ability
Technical 2 1.65 3.3 2 1.65 3.3 2 1.8 3.6
performance
Functional 2 1.1 2.5 2 −0.75 −1.5 2 1 2
effectiveness
(W) = 7 7 7
(W * Score) = 4.55 −1.45 10.85
(Weighted average 0.6 −0.2 1.6
score) ITech =
Note On weighted average score: −2 to −1 = very bad; −1 to 0 = bad; 0 = neutral; 0 to +1 = good; +1 to +2 = very good
N. Nowshin et al.
Local labor and local inexpensive material were used for the construction of
the retrofitted house. The weighted average score for this technology is +1.6. The
design and construction of this technology were much simpler compared to those
of the amphibious units. The wooden floor floated perfectly with a reasonable load.
Minimal obstructions were created during the vertical movement of the floor, which
settled on the ground after floodwater drained out. No object was trapped under the
floor. No major technical difficulties are reported for this technology. Minor technical
problems occurred during the last two flood seasons, which were fixed by the users
themselves. However, the users do not like the wooden floor, because it is visually
unpleasant to them. They also reported infestation of insects inside the floor and a
low load-bearing capacity of the floating base.
Table 4 gives the weighted average scores for the economic indicators of sustain-
ability. The amphibious living unit has an economic score of +0.7 which is satis-
factory/good. The users had to lose the land where they grew vegetables for extra
income and household consumption. They used to sell their extra vegetables for
around BDT 6,000 per year. Additionally, this technology is relatively expensive and
not very suitable for the poor communities due to its high construction cost. On the
other hand, this living unit saved the users from loss of household assets during the
floods. The users moved all their valuables and perishable assets including cattle,
looming machine, etc., to the amphibious house. The users are satisfied with the fact
that they could operate their thread machine during floods, which is their primary
source of livelihood.
Although the amphibious sanitation unit is constructed on the same land, the
weighted average (economic) score for this technology is much lower (−0.6). The
sanitation unit not only required a relatively high unit cost, but also high maintenance
and repair costs, particularly during floods.
The economic weighted average score for the retrofitted house is +1.1. This
score is in the range “very good” and results from the low construction cost as well
as low maintenance and repair costs. The low cost is particularly appreciated by the
poor flood effected community. However, the users were unable to keep their heavy
looming machine on the wooden floor to continue their livelihood activities for about
1.5 months. The user used the floor as a storage space so no household items were
damaged.
Although the amphibious living unit is quite expensive, this technology scores
the highest (+1.305) in terms of social indicators of sustainability (Table 5). The
reason behind the remarkable social acceptance is the high load-bearing capacity and
technological flexibility; that is, it can be moved easily from one place to another
during flood and can be set perfectly on the ground after flood. Moreover, this unit can
be used in the charlands, flooded areas and water logged areas. The local community
is very much interested in this technology. However, they would like to have this
technology free of cost. Neighbors and relatives of the users stayed in this house and
helped the users in the maintenance works. The only problem was the scarcity of
drinking water. Since they were not affected by any waterborne diseases, they did not
want to spend any money for water treatment. Various visitors, e.g., NGOs, LGED
82
Table 4 Weighted average scores for the economic indicators of sustainability

Return on 1 −0.25 −0.25 1 −1.5 −1.5 1 0.75 0.75
investment
Business agility 1 −2 −2 1 −2 −2 1 1.5 1.5
Life-cycle cost 1 −1 −1 1 0.25 0.25 1 2 2
Willingness to pay 2 2 4 2 0.875 1.75 2 0.5 1
Employment 1 1.25 1.25 1 0.75 0.75 1 1.5 1.5
Total capital cost 3 0.5 1.5 3 −2 −6 3 1 3
Construction time 2 2 4 2 0 0 2 1.375 2.75
(W) = 11 11 11
(W * Score) = 7.5 −6.75 12.5
(Weighted average 0.7 −0.6 1.1
score) IEco =
Note on weighted average score: −2 to −1 = very bad; −1 to 0 = bad; 0 = neutral; 0 to +1 = good; +1 to +2 = very good
N. Nowshin et al.
Table 5 Weighted average scores for the social indicators of sustainability
Labor practice and 1 1.5 1.5 1 0.5 0.5 1 2 2
decent work
Society and customers 1 1.6 1.6 1 0.75 0.75 1 0.5 0.5
Human health risk 1 0 0 1 0.5 0.5 1 −1 −1
impact
Acceptability to 2 0.5 1 2 0 0 2 1 2
stakeholders
Participation and 1 2 2 1 2 2 1 0.25 0.25
responsibility of
different stakeholders
Wellbeing/happiness 1 1.25 1.25 1 0.5 0.5 1 2 2
Quality of life 2 1.5 3 2 0.5 1 2 1.25 2.5
Co-creation of Flood Mitigation Technologies in Bangladesh …
Flexibility and 2 2 4 2 2 4 2 0.5 1

adaptation
(W) = 11 11 11
(W * Score) = 14.35 9.25 9.25
(Weighted average 1.3 0.8 0.8
score) ISoc =
83
and people from faraway places, are increasingly coming to visit the living unit on
a regular basis.
Although used by the same users, at the same location and having a similar
technology, the amphibious sanitation unit scores +0.8 because of the difficulties and
challenges in maintenance and repair due to the unavailability of expert electricians
and plumbers locally. Visitors are curious also about this technology and appreciate
this innovative technology and its usefulness. Public consultation indicates that the
main social barriers for using this technology include congested toilet space, lack of
privacy and malfunctioning of sewerage. In general, women are hesitant to use this
unit, as it is placed at a higher ground and can be visible from a distance. Neighbors
and the local community are very much interested in this technology, but they want
this technology free of cost too. The users are happy to have this unit. However, they
are not too comfortable using this unit as they are not completely satisfied with the
use of the flushing mechanism and other unfamiliar systems. The toilet unit became
electrified when rainwater entered the toilet unit. So they could not use the toilet unit
when it was raining and was forced to use the regular toilets.
The retrofitted house scores +0.8 in terms of the social dimension. The users
are happy to have this technology because at least they could save their household
assets independently without bothering their neighbors. Their only demand is to
strengthen this floor for more durability so they could keep more weight along with
their looming/thread machine. However, the local community, neighbors, manu-
facturers or any government official/organization did not show any interest in this
technology. Additionally, the user could not stay in this house during the last flood
year (2020) because there is no attached toilet facility, and it is more difficult to
access.
The amphibious living unit is located in an open environment that used to have
many trees which attracted birds. Additionally, during the flood period, this location
was the habitat for fish. Moreover, the users cultivated rice and vegetables during
dry season. Construction of the amphibious unit changed this natural environment
significantly. Although there is no direct wastewater discharge from the living unit,
dumping of household wastes in the floodwater causes pollution. As a result of these
considerations that followed a public consultation, the weighted average score of the
environmental sustainability indicators is −0.3 (Table 6).
On the other hand, although the sanitation unit was constructed in the same loca-
tion and caused the same environmental changes, the environmental sustainability
score for this unit is +0.2. This higher score is due to the fact that the users and other
respondents are satisfied with the way the wastewater discharge and environmental
impacts are managed in this unit.
Construction of the retrofitted house did not cause any damage to trees or crops.
The construction only modified the floor of an existing house. The main construction
material is wood which is natural and environment friendly. The recycled plastic
bottles used in the floor platform improved environmental pollution. However, there
was some unpleasant odor emitting from the floor, specifically after the floodwater
drains out. The users report infestation of the floor with insects and cockroaches.
Table 6 Weighted average scores for the environmental indicators of sustainability
Impact on 2 −2 −4 2 −2 −4 2 2 4
environment
(positive/negative)
Waste (prevention of 1 −2 −2 1 2 2 1 −1.5 −1.5
waste, recycling,
disposal)
Material and resource 1 2 2 1 1.5 1.5 1 1.75 1.75
(reusability and
waste)
Pollution (land and 2 2 4 2 2 4 2 −1 −2
water)
Impact of land use in 1 −2 −2 1 −2 −2 1 1 1
ecosystem
(W) = 7 7 7
(W * Score) = −2 1.5 3.25
(Weighted average −0.3 0.2 0.5
score) IEnv =
85
Considering these situations, the environmental sustainability score for the retrofitted
house is +0.5.
4.2 Comparative Sustainability Analysis
Following the unranked pair-wise comparison method (Canter 1996), relative weights
are assigned to each dimension of sustainability to calculate the combined sustain-
ability score. The relative weights together define the overall sustainability score. In
this study, for sustainability considerations, the technical performance was found to
have preference over cost. For example, between two technology options, the option
with a better technical performance could be selected to be more sustainable even if
it is costlier. This results from the fact that the respondents placed higher importance
(relative weight) on the technical indicators than the economic indicators.
Similarly, when two options or technologies have the same levels of technical and
environmental performance, the option with higher economic advantage (cheaper)
is selected even it has less social acceptability.
This concept is used in unranked pair-wise comparison method (Table 7) to fix
the relative weights for different dimensions. This method compares the importance
of a dimension/factor in relation to each of the other dimensions/factors. Total 10
pairs of comparison can be made among these 5 dimension/factors (4 dimensions
+ 1 dummy factor) because 5 C2 = 10. A weight of “1” indicates more important
dimension/factor of a pair and “0” indicates a less important one. In this method,
a factor-importance coefficient (FIC) of a factor or dimension is equal to the sum
of all pair-wise assigned weights (S) divided by the sum of all weights (T1) for all
dimensions/factors. A dummy factor is defined as that factor of each pair, deemed
less important of the two, and acts as a “place keeper” to avoid skewness of the
process (Canter 1996).
The technical dimension or the importance of sustainable flood resilience tech-
nology is considered the most important for sustainability and is assigned the highest
weight, i.e., 40%. Long-term economic viability or the economic dimension is the
next important dimension and is assigned a relative weight of 30%. This means that
a technically feasible technology even with sound performance may not sustain if
the long-term economic viability is very poor. With similar comparative consider-
ations, 20% and 10% relative weights are assigned to the social and environmental
dimensions of sustainability, respectively.
Finally, the combined sustainability scores for the three technologies are calcu-
lated by aggregating the sustainability scores (= relative weight * weighted average
scores in Tables 3–6) determined for the four dimensions of sustainability (see
Table 8). The retrofitted house is the most sustainable technology (score = +1.2),
followed by the amphibious house (score = +0.7) and the sanitation unit (score =
−0.1). The negative score for the sanitation unit indicates unsustainability of the
technology.
Table 7 Unranked pair-wise comparison method for importance weight assignment
Factor/dimension Assignment of weight Row-wise sum Factor-importance coefficient FIC) = Relative weight (RV) = (FIC*100)
(S) (S)/T1 (%)
Social 0 0 1 1 2 0.20 20
Technical 1 1 1 1 4 0.40 40
Economic 1 0 1 1 3 0.30 30
Environmental 0 0 0 1 1 0.10 10
Dummy 0 0 0 0 0 0.00 0
Column total T1 = 10 1 100
Note 1 = more important dimension/factor; 0 = less important dimension/factor

87
88
Table 8 Combined sustainability scores

Weighted average score Sustainability score
Dimensions Relative weight Amphibious living Amphibious Retrofitted house Amphibious living Amphibious Retrofitted house
(RV) (%) unit sanitation unit unit sanitation unit
Technical 40 0.6 −0.2 1.6 0.243 −0.083 0.620
Economic 30 0.7 −0.6 1.1 0.205 −0.184 0.341
Social 20 1.3 0.8 0.8 0.261 0.168 0.168
Environmental 10 −0.3 0.2 0.5 −0.029 0.021 0.046
Combined Sustainability Score 0.7 −0.1 1.2
N. Nowshin et al.
4.3 Expert Opinion
Among different stakeholders, the technical and socio-economic experts provided

very useful insights regarding the sustainability of these technologies. The experts
agreed with the location selection and user group selection as well. As per the experts,
these technologies are constructed in the safe zone. The real sustainability situation
would be understood if these technologies faced extreme geographic or climatic
conditions like near erosion-prone river banks, flash flood, cyclone, storm surge,
strong wave, high tides, etc. The experts also expressed their concern over using costly
and not easily obtainable foreign materials in the construction of amphibious units.
They emphasized on inclusive design: local materials and feedback from local stake-
holders. Technology transfer for local production can be initiated through govern-
ment interventions (i.e., Dutch embassy, etc.). According to the experts, the toilet
unit is the technology that is most in demand as it prevents pollution. But, it lacks
easy access for elderly and other groups, and hence, some modifications regarding
its capacity and access are suggested. As retrofitted house is constructed with local
material, this technology is more feasible in terms of sustainability. For retrofitted
house, reuse of plastic bottle is quite appreciable, as it reduces the problem of solid
wastes. These plastic bottles are not exposed in direct sunlight. Therefore, no UV
radiation occurred and no harmful elements were produced by the products. In addi-
tion to the introduction of these technologies, alternate employment opportunities
need to be created simultaneously with existing livelihood through locally available
capacity building programs. If the user group wants to avoid moving to shelter houses,
they can build floating gardens, take on fish farming, etc., to keep them employed
throughout the year. They can even construct a floating storage area to keep their
agricultural production safe during floods.
4.4 Policy Recommendations
The following policy recommendations are made based on public consultation and
the overall analysis and findings.
1. Any new intervention should be built based on the knowledge and technology
that exists at the local level. This would ensure ownership and sustainability of
the new technologies.
2. It is important to ensure participation of the local community of users, policy
makers, skilled workers, and other stakeholders in the process of conceptualizing
and constructing the pilot interventions. This would help develop ownership of
the interventions and promote scale up of the technologies. Introducing ‘on the
job’ technical education and training, investing in infrastructure that enables
communities, and including access to information services, data, and the internet
would help in this circumstance.
3. To ensure the sustainability of a technology, it is essential to understand the

technical, economic, social and environmental complexities of improving and
scaling up the technology in the backdrop of cultural and behavioral dynamics
of the local context.
4. Needs’ assessments would be conducted in consideration of social and political
acceptability of changes, and incentives needed for behavioral changes.
5. BDP2100 should foster the inclusion of local level innovation in terms of ideas,
regulations, governances and institutional reforms, stakeholder engagement,
techniques and technologies; avoiding the trap of expecting conventional, large-
scale engineering approaches to solely be able to cope with the future threats
and ever changing demands.
5 Conclusion
This study conducted focus group discussions, workshops, meetings and interviewed
the users, manufacturers and other local stakeholders in a peri-urban area of Siraj-
ganj, Bangladesh, to assess the sustainability of some of the small-scale flood miti-
gation technologies that have been implemented. The combined sustainability of a
technology is assessed by evaluating the indicators of the four dimensions of sustain-
ability, i.e., technical, economic, social and environmental. Of the three technologies
assessed, i.e., amphibious house, amphibious sanitation unit and retrofitted house,
the retrofitted house is found to be the most sustainable technology. Based on public
consultation, field observation and analysis of the collected data and information, it is
concluded that successful technical performance and functional effectiveness are the
primary requirement for sustainability of a flood mitigation technology to improve
community resilience. In the technology adoption process, it is important to follow a
co-creation approach where the stakeholders are engaged early on in the process so
they have more ownership of the technology, are willing to pay for the technology, and
are willing to co-invest to scale up and replicate the technology. Economic sustain-
ability is the second requirement, which defines the communities’ ability to afford the
adoption of the technology. Social acceptability is another requirement which defines
the quality of life, flexibility and adaptive capacity of the flood-affected community.
The fourth, environmental sustainability, is important to assess the implementation
of measures in relation to all species, habitats and landscapes.
It is essential to engage the local communities and relevant stakeholders including
the local mechanics, small businesses and technical education institutions to ensure
uptake, replication and scale up of the technology with appropriate modification in
the local context.
In summary, this study suggests that efforts to boost community resilience and
adaptive capacity would benefit from the sustainable development of flood miti-
gation technologies considering the technical, economic, social and environmental
dimensions, and involvement of all relevant stakeholders in all stages of the process.
Acknowledgements Funding for the research leading to this article is provided by Netherlands
Organization for Scientific Research, The Netherlands.
References
Brammer, Hugh. 1990. Floods in Bangladesh: Geographical Background to the 1987 and 1988
Floods. The Geographical Journal 156 (1): 12–22. https://doi.org/10.2307/635431.
Brammer, Hugh. 2004. Can Bangladesh Be Protected from Floods? Dhaka: The University Press
Limited.
Canter, Larry W. 1996. Environmental Impact Assessment. Singapore: McGraw-Hill.
CORE Bangladesh Project. 2016. NWO. Accessed 19 May 2021. https://www.nwo.nl/en/projects/
w-0769203.
Corsel, Denise. 2015. Addressing Worldwide Flood Concerns: Empowering Local Commu-
nities. MEI@75, June 15. https://www.mei.edu/publications/addressing-worldwide-flood-con
cerns-empowering-local-communities.
Dasgupta, Susmita, Mainul Huq, Zahirul Huq Khan, Md. Sohel Masud, Manjur Murshed Zahid
Ahmed, Nandan Mujharjee, and Kiran Panday. 2011. Climate Proofing Infrastructure in
Bangladesh: The Incremental Cost of Limiting Future Flood Damage. The Journal of Envi-
ronment & Development 20 (2): 167–190. https://journals.sagepub.com/doi/10.1177/107049651
1408401.
Deubelli, Teresa M. 2019. Flood Resilience Measurement for Communities: Boosting Community
Resilience. Prevention Web, May 16. https://www.preventionweb.net/news/view/6549.
Dewan, A.M., Makoto Nishigaki, and Mitsuru Komatsu. 2003. Floods in Bangladesh: A Compar-
ative Hydrological Investigation on Two Catastrophic Events. Journal of the Faculty of Environ-
mental Science and Technology, Okayama University 8 (1): 53–62. http://ousar.lib.okayama-u.
ac.jp/files/public/1/11503/20160527190938720714/008_053_062.pdf.
Edjossan-Sossou, A.M., O. Deck, M. Al Heib, and T. Verdel. 2014. A Decision-Support Method-
ology for Assessing the Sustainability of Natural Risk Management Strategies in Urban Areas.
Natural Hazards and Earth System Sciences 14 (12): 3207–3230. https://doi.org/10.5194/nhess-
14-3207-2014.
Elsheikh, Ranya Fadlalla Abdalla, Sarra Ouerghi, and Abdel Rahim Elhag. 2015. Flood Risk Map
Based on GIS, and Multi Criteria Techniques (Case Study Terengganu Malaysia). Journal of
Geographic Information System 7 (4): 348–357. https://doi.org/10.4236/jgis.2015.74027.
Flower, Ben, and Matt Fortnam. 2015. Urbanising Disaster Risk. Cambodia: People in
Need Mission. https://www.preventionweb.net/files/submissions/47109_urbanisingdisasterris
kreportinteractive.pdf.
Goodman, Jennifer, Angelina Korsunova, and Minna Halme. 2017. Our Collaborative Future: Activ-
ities and Roles of Stakeholders in Sustainability-Oriented Innovation. Business Strategy and the
Environment 26 (6):731–753. https://doi.org/10.1002/bse.1941.
Islam, Mohammad N. 2017. Living with Hazards. Dhaka: University Grants Commission of
Bangladesh.
Kabir, Md. Humayain, and Md. Nazmul Hossen. 2019. Impacts of Flood And Its Possible Solution
in Bangladesh. Disaster Advances 12 (10): 48–57. https://www.researchgate.net/publication/336
146425.
Keating, Adriana, Karen Campbell, Reinhard Mechler, Erwann Michel-Kerjan, Junko Mochizuki,
Howard Kunreuther, JoAnne Bayer, Susanne Hanger, Ian McCallum, Linda See, Keith Williges,
Ajita Atreya, Wouter Botzen, Ben Collier, Jeff Czajkowski, Stefan Hochrainer, and Callahan
Egan. 2014. Operationalizing Resilience Against Natural Disaster Risk: Opportunities, Barriers,
and a Way Forward. Zurich Flood Resilience Alliance. http://pure.iiasa.ac.at/id/eprint/11191/1/
zurichfloodresiliencealliance_ResilienceWhitePaper_2014.pdf.
Mirza, M. Monirul Qader. 2003. Three Recent Extreme Floods in Bangladesh: A Hydro-
Meteorological Analysis. Natural Hazards 28: 35–64. https://doi.org/10.1023/A:102116973
1325.
Montz, Burrell E., and Eve Gruntfest. 2002. Flash Flood Mitigation: Recommendations for Research
and Applications. Global Environmental Change Part B: Environmental Hazards 4 (1): 15–22.
https://doi.org/10.3763/ehaz.2002.0402.
Parvin, Gulsan Ara, Annya Chanda Shimi, Rajib Shaw, and Chaitee Biswas. 2016. Flood in a
Changing Climate: The Impact on Livelihood and How the Rural Poor Cope in Bangladesh.
Climate 4 (4): 60. https://doi.org/10.3390/cli4040060.
Rouf, Tasnuva. 2015. Flood Inundation Map of Sirajganj District Using Mathematical Model.
MSc. Thesis, WRE, BUET. Accessed 19 May 2021. http://lib.buet.ac.bd:8080/xmlui/handle/123
456789/966.
Sims, Lynn O., Robert C. Knox, and Larry W. Canter. 1991. Protocol for Selecting an In-Tank Leak
Detection Device. International Journal of Environmental Studies 39 (3): 163–176. https://doi.
org/10.1080/00207239108710692.
Sinthumule, Ndidzulafhi I., and Ntavheleni V. Mudau. 2019. Participatory Approach to Flood
Disaster Management in Thohoyandou. Jàmbá: Journal of Disaster Risk Studies 11 (3): a711.
https://doi.org/10.4102/jamba.v11i3.711.
Smith, Adrian, Mariano Fressoli, Dinesh Abnol, Elisa Arond, and Adrian Ely. 2016. Grassroots
Innovation Movements. London and New York: Routledge. https://www.taylorfrancis.com/
books/mono/10.4324/9781315697888/grassroots-innovation-movements-adrian-smith-mar
iano-fressoli-dinesh-abrol-elisa-arond-adrian-ely.
Sultana, Parvin, and Paul Thompson. 2019. Livelihoods in Bangladesh Floodplains. In Oxford
Research Encyclopedia of Natural Hazard Science. New York: Oxford University Press. Accessed
19 May 2021. https://doi.org/10.1093/acrefore/9780199389407.013.258.
UNDRR. 2020. Dramatic Rise in Climate over Last Twenty Years. UN Report. Accessed 21 June
2011. https://www.preventionweb.net/news/view/74121.
Urban Waterlogging Risk Profiling: The
Case of Khatunganj Wholesale
Commodity Market, Chattogram
Tasnim Alam Nishat, Dewan Salman Sunny, Rifat Talha Khan,

Md.Reaz Akter Mullick, and Piyal Datta
Abstract Khatunganj, a busy trade region extending across an area along the Chaktai
canal in Chattogram, has been severely affected by recurrent inundation for the last
few years. The presented study used a reproducible methodology to develop a risk
and vulnerability profile of Khatunganj connecting waterlogging with climate change
scenario. For hydrologic-hydraulic modeling purpose, a model was set up on SWMM
numerical modeling platform to simulate the discharge and water level at different
points of the canal. Nine different scenarios of rainfall events were considered, where
one actual rainfall event was taken, and the remaining scenarios were considered as
design rainfall events. Using an inundation pattern combined with three other factors,
namely, proximity to canal, structure type, and land use type of the area, the vulner-
ability of the area was assessed. Vulnerability of 224 households was calculated for
low, average, and medium rainfall using the Household Sector Approach in ArcGIS.
Analyses of the nine scenarios showed that even a 50% decrease in rainfall will
generate a water level of 3.09 m to 3.84 m, whereas the elevation level of Khatunganj
main road is between 2.9 m and 3.1 m. Results found that 27, 30, and 71 structures
are very vulnerable at the time of low, medium, and high rainfall, respectively.
Keywords Khatunganj Commercial Area · Design rainfall · Flood inundation ·

Vulnerability · Risk profile
1 Introduction
Urban waterlogging refers to the phenomenon when a short duration intense rain-
fall event surpasses the urban drainage capacity and causes an adverse waterlogged
situation (Xue et al. 2016). Rapid urbanization accelerates the change of urban land
surfaces, typically by diminishing ecological land and expanding the impenetrable
cover regions, which keeps water from penetrating into the soil (Deng et al. 2020).
T. A. Nishat · D. S. Sunny · R. T. Khan · M. A. Mullick (B) · P. Datta

Department of Civil Engineering, Chittagong University of Engineering and Technology,
Chittagong, Bangladesh

94 T. A. Nishat et al.
Another reason for urban waterlogging is heavy shower in urban areas, which is
a result of intensified heat island effects occurring in recent years (Miller et al.
2014; Alexander et al. 2019). Most climate models and empirical evidence in the
twenty-first century demonstrate that as the climate gets warmer, atmospheric water
vapor increases, resulting in more intense and frequent precipitation events, which are
likely to exacerbate the current drainage congestion scenario (Berndtsson et al. 2019).
Following heavy rain showers, flash floods on urban streets and protracted flooding
of low-lying areas have become a pervasive problem in various cities of Bangladesh,
particularly in the port city and commercial capital of Chattogram. Chattogram makes
significant contributions to the national economy, generating 40% of the country’s
industrial output, 80% of its international trade, and 60% of government revenue (E.
Kabir 2016).
Khatunganj is the primary wholesale market for daily commodities in Chattogram
and is vital not only for the economy of the city, but also for the entire national
economy (Chowdhury 2011). Khatunganj was originally developed to serve as a
business hub as it is situated at the bank of Chaktai canal, which was the gateway
to inner land from the Chattogram Port. However, Khatunganj has been subject to
severe waterlogging problem for about one and a half decades till date. Khatun-
ganj traders reported suffering from excruciating costs due to waterlogging, as their
storage of raw commodities is often destroyed. A very short duration rainfall results in
waterlogging in all the alleys of this crowded market area (Quddusi 2017). Traders in
Khatunganj suffered direct financial losses of around $60 million last year solely due
to waterlogging (Barua 2021). The waterlogging problem at this wholesale market
appears to have barely been resolved by the city managers and policy makers. Very
little has been done to conduct a scientific study to assess the economic impacts of
the current waterlogging scenario.
Researchers have introduced a range of different risk assessment methodolo-
gies for waterlogging. Peng-zhu et al. (2000) produced Hydrologic and Hydraulic
(H&H) models for different precipitation frequencies and showed flood inundated
conditions with geomorphologic classification mapping for the Lake Taihu region
in China. The Hydrologic and Hydraulic model (H&H) in the watershed catchment
is used to simulate flood, and to thereby evaluate the impacts of urban develop-
ment and the effectiveness of watershed management (Nayeb Yazdi et al. 2019). The
Storm Water Management Model (SWMM) is widely used in H&H in urban areas
because it has shown the capability to simulate conveyance systems even in very
small-scale (2 hectares) urban contexts (Nayeb Yazdi et al. 2019; Wanniarachchi
and Wijesekera 2012). Jang et al. (2007) explored the suitability of SWMM as an
urban hydrology modeling tool for post-development scenario development. Using
hydrological modeling and GIS spatial analysis, Quan (2014) created a paradigm
for waterlogging risk assessment under six hypothetical scenarios. The concept of
multi-scenario analysis has piqued interest in waterlogging research in recent years.
Its objective is to evaluate the different risks posed by a hazard, as well as the indi-
vidual hazard intensity assigned to each aspect (Shi et al. 2010). For the Jing’an
District of Shanghai, China, Yin et al. (2011) performed a scenario simulation to
evaluate the risk of small-scale urban waterlogging disasters across various return
Urban Waterlogging Risk … 95
periods. Several other studies have applied the scenario analysis method to evaluate
the risk of waterlogging (Dutta et al. 2003; Oliveri and Santoro 2000; Kleist et al.
2006). Several studies also argued that climate change has been a contributing factor
to flood risk by raising the precipitation amount much higher than the average annual
rainfall (Fleming et al. 2012; Hirabayashi et al. 2010). However, the severity of water-
logging hazard is only one of the factors that determine water risk and vulnerability.
Other variables such as exposure play a part in risk profiling as well. When flooded
land does not contain any components, there are no risks since zero exposure equals
zero risk. Therefore, there is a need to evaluate exposure by modeling inundation
degree for different scenarios (Shi et al. 2010). Han et al. (2006) carried out a research,
where road exposure was evaluated using scenario simulation.
It is evident that the entire Chattogram city is faced with the problem of water-
logging every year. But a long-term solution for the overall city will only prolong
the sufferings. Rather, an integrated solution based on a priority area can be effective
and economically beneficial in the short term. Since the market at Khatunganj is
of the wholesale kind, the stock is massive in general, and businessmen hardly get
any time to transfer their products to a safe custody at the onset of waterlogging.
The retail price of daily products in and around Chattogram city increases as an
indirect effect. However, an intensive study on this prime location can prevent the
loss of millions of dollars every year. Therefore, this study attempts to provide an
answer to the research question: ‘What is the extent of the inundation problem due
to different rainfall scenarios and what would be the risk map of Khatunganj due to
urban flooding considering climate change?’. Three objectives have been formulated
to answer the research question-
1. To develop Hydrologic and Hydraulic model to simulate discharge and water
level at Khatunganj from various rainfall and tidal scenarios.
2. Develop a methodology for waterlogging risk and vulnerability assessment with
climate change scenarios and
3. To develop a risk and vulnerability profile of Khatunganj in relation to
waterlogging with climate change scenario.
2 Study Area
Khatunganj is situated in the eastern part of the Chattogram city, alongside the famous
Chaktai canal. This is the location where the Chaktai canal opens to the Karnafuli
River. The geographic location of Khatunganj is at 22.330N and 91.840E. It is located
at Ward No. 35 of Chattogram City Corporation. It is a relatively low-lying area within
the city, situated only 3 m above the sea level. The temperature in the area has been
found to vary from 14.5 to 39.3 ˚C during the monsoon season over the period of
1950–2002 (Nasher and Uddin 2015). Chattogram experienced a mean rainfall of
16 mm at dry season, 444 mm at pre-monsoon, 1857 mm at monsoon, and 267 mm at
post-monsoon, for the period of 1949 to 2011 (Hasan et al. 2014). The city is located
beside the Karnafuli River which is semi-diurnal in nature. Tidal ranges from chart
datum (ISLWL-Indian Spring Low Water Level which is 1.673 m below mean sea
level) are 1.5 m–5.5 m (above ISLW) at Patenga, 1.5 m–4.8 m (above ISLWL) at
Khal No. 10, and 1.2 m–4.2 m (above ISLW) at Sadarghat (S. M. I. Kabir and Ali
2017). Khatunganj is located at the downstream part of Murarpur-Bahaddarhat sub-
catchment. The whole area becomes flooded following a period of extensive rainfall,
as Chaktai canal cannot hold the water flowing from upstream (Fig. 1).
(a) (b)
(d) (c)
Fig. 1 a Bangladesh within world map; b Location of Chattogram within Bangladesh; c Chattogram
City Corporation; d Location map of Khatunganj and Chaktai Canal in Chattogram
3 Methodology
3.1 Preparation of Digital Elevation Model (DEM)
Hydrologic processes and water resource issues are commonly investigated by using
distributed watershed models. These watershed models require physiographic infor-
mation such as configuration of the channel network, location of drainage divides,
channel length and slope, and sub-catchment geometric properties. These param-
eters are usually gathered from secondary maps or field surveys. During the last
two decades, this information is generally acquired directly from digital topo-
graphic representation, which is called ‘Digital Elevation Model (DEM)’ (Jenson
and Domingue 1988).
Digital data generated by this approach also has the advantage that Geographic
Information Systems (GIS) can easily import and analyze these data. The technical
advancements made by GIS and the growing availability and quality of DEMs have
greatly expanded the scope for application of DEMs to many hydrologic, hydraulic,
water resources, and environmental research (Jenson and Domingue 1988).
Since the study site is not very large, a fine-scale Light Detection And Ranging
(LiDAR)-generated DEM, preferably 0.5 m × 0.5 m, was required and was used for
this study. ArcGIS 10.8 and ArcScene 10.8 were used to prepare and visualize the
Digital Elevation Model (DEM), respectively.
3.2 Preparation of Digital Surface Model (DSM)
Digital Elevation Models (DEMs) have diverse and well-documented applications

including visual impact assessment, hydrological modeling, flood prediction, and
site suitability analysis. Automated elevation model creation from remotely sensed
data may provide a representation of both the ground surface and the objects on that
surface (Burrough et al. 1998). Digital Surface Models (DSMs) provide the option of
extracting surface elevations to leave the surface DEM on the ground. This separation
of over-surface feature information from ground information can provide a useful
combination of data sets for many applications in the urban world. For example,
comprehensive knowledge of ground surface elevation is necessary to predict flood
inundation and the possible effects of sea-level rise, whereas a comprehensive model
of man-made structures that will be impacted as a result is important for land owners,
planning authorities, and insurance companies. Since the study site is not very large,
a fine-scale LiDAR-generated DSM, preferably 0.5 m × 0.5 m, was used. ArcGIS
10.8 was used to produce Digital Surface Model (DSM), and ArcScene 10.8 was
used to visualize it.
3.3 Watershed Analysis
Watershed or catchment delineation was carried out for the entire basin containing
the Khatunganj sub-catchment. A watershed is the land area that drains water to the
outlet (Chaktai canal) during a rainstorm. It is simply an area that drains surface water
from high elevation to low elevation. For delineation of the catchment boundaries,
at first, the drainage network was identified from field visits to the study site and its
surrounding areas. The drainage network was then drawn in ArcGIS by manipulating
the Aster 30 m resolution DEM using hydrological tools in ArcGIS. DEM filling
function was performed to avoid irregular stream/drainage networks. For showing
the direction of water flows out of each cell of a filled elevation raster, the flow
direction tool was used. The flow accumulation tool was employed to tabulate the
accumulated flow as the accumulated weight of all cells flowing into each downslope
cell in the output raster. Stream networks were derived from a flow accumulation
raster. Finally, raster data was converted to vector data and overlaid on satellite image
for verification.
3.4 Hydrologic and Hydraulic Modeling
The Storm Water Management Model (SWMM), established by the US Environ-

mental Protection Agency (EPA), is used to set up hydraulic models. The model
simulates hydrologic response in dendritic watershed. A hydrological model can be
described as a set of equations, which help to estimate runoff as a function of the
various parameters used to describe the characteristics of the watershed. To esti-
mate the extent and depth of inundation due to a specific storm, hydrologic and
hydraulic modeling is required. Rainfall, basin characteristics, infiltration losses,
calibration data, and different human effects are regarded as the data required for
hydrological modeling. Cross section of the Chaktai canal at 89 points (Fig. 2),
water level of Chaktai outfall for the year 2017, and discharge of Chaktai outfall for
3 months (June, July, and August) of 1983 were collected from Centre for Environ-
ment and Geographic Information Service (CEGIS). Rainfall data (1980–2017) for
Chattogram was obtained from the Meteorological department (BMD).
To simulate the rainfall-runoff scenario, the Soil Conservation Service (SCS)
curve number method was used for infiltration loss. Curve number of each basin
was collected from Mishra and Singh (Mishra and Singh 2013) according to the
relevant soil group (as presented in Fig. 3) and imperviousness (Fig. 4). The SCS
unit hydrograph has been used to transform precipitation excess to direct run-off.
In this regard, lag time (min) was required for the transform method and following
equation was used to calculate the lag time (Hour),
0.7
2.587 × L 0.8 × 1000 −9
Lag time(h) = CN
(1)
1900 × H 0.5
Fig. 2 Muradpur- Bahaddarhat basin with sub-catchments and cross-section points of Chaktai khal
90
Curve number
85
80
75
70
S S S S S S S S S S S S S S S S S S S S S S S S S S S S S S S S
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
Sub-basin name
Fig. 3 Curve number of Muradpur-Chawkbazar basin
100
Impervious (%)
80
60
40
20
0
S S S S S S S S S S S S S S S S S S S S S S S S S S S S S S S S
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
Sub-basin name
Fig. 4 % Imperviousness of Muradpur-Chawkbazar basin

Where, L = Hydraulic watershed length in meter = 110 × A0.6 ;

A = Sub-catchment area in hectare;
CN = Curve number;
H = Average sub-catchment land slope in percentage.
Average sub-catchment land slope was calculated from DEM using ArcGIS inter-
face. Due to lack of data availability, base flow was not used. Since most of the areas in
the basins are urbanized and impervious, interception and surface method were also
not used. Since the durations of the rainfall and resulting runoff events are relatively
short, and the events mostly happen in the rainy season with highly humid conditions,
evaporation losses are very low and are considered to be negligible (Schwab et al.
1993).
The model was run with an actual rainfall event i.e., the maximum rainfall event
in 2017 on 05 July, and also simulated with a number of design rainfall events. The
Rainfall Intensity-Duration-Frequency (IDF) curve has been used in this study to
create the design of rainfall events. IDF curves graphically represent the amount of
water that falls in a catchment within a given period of time. The following IDF
relationship (Fig. 5) was used to extract the design rainfall events. IDF equation is
based on the rainfall for the years 1980–2017 in Chattogram and follows the following
equation.
776.24 × T r 0.217
Rainfall intensity(mm/hr.) = (2)
T d 0.563
where,
Tr = Return period in year;
Td = Rainfall event duration in hour.
Fig. 5 IDF Curve of Chattogram City by Hershfield Method

Several scenarios with different critical design storms based on IDF analysis
were generated using equations obtained from literature. The following scenarios
have been considered:
1. Scenario 1: Max rainfall event in 2017 distributed over 24 h (Actual rainfall
event scenario)
2. Scenario 1 (A): Max rainfall event in 2017 (2-h, High Tide)
3. Scenario 2: 2-h 20-year return period rainfall (High Tide)
5. Scenario 4: 2-h 50-year return period rainfall (Low Tide)
6. Scenario 5: 2-h 50-year return period rainfall (considering 50% rainfall
decrease) (High Tide)
7. Scenario 6: 2-h 50-year return period rainfall (considering 50% rainfall
decrease) (Low Tide)
9. Scenario 8: 2-h 50-year return period rainfall (considering Climate change 11%
rainfall increment)
The resulting runoff from the hydrologic models of various sub-catchments was
routed through the nodes, and conduits were added in the SWMM model. The cross-
section properties of the conduits were selected as irregular shaped, and relevant
data obtained from the bathymetric survey conducted by CEGIS was given as input.
Dynamic wave routing and unsteady flow equations were used in simulating the
model. The Manning’s roughness coefficient, n-value was taken as 0.01. At the
outlet, the boundary condition was set as 4.25 m water level (WL), based on the
highest tide WL in the year 2017 due to water level data being available only for that
year. To incorporate the scenario of climate change-induced sea-level rise, a 0.60 m
rise was estimated (Ministry of Foreign Affairs of the Netherlands 2019), and 4.85 m
WL was used to include the climate change scenario.
The entire catchment has been considered for hydrological modeling, and no
boundary condition was imposed to the model. However, water level at the outfall of
Chaktai canal has been used as the downstream boundary condition. The high and
low tide condition water levels at the Chaktai outfall are considered as maximum and
minimum boundary level. The WL is taken as 4.5 m and 1.5 m for high and low tide
conditions, respectively, based on the tide chart published by the Chittagong Port
Authority.
3.5 Inundation Maps
Inundation scenarios were illustrated using the simulation tool of ArcScene 10.8.
DEM of a rectangular area covering Khatunganj was taken in ArcScene. ‘Floating on
a custom surface’ was activated at ‘Base Height’ in ‘Properties’ of Arcscene. A blue
color polygon was assigned to represent the inundation. At ‘Animation Manager’,
a total of 13 keyframes were taken, out of which 11 keyframes represent the 11
inundation scenarios. The other two keyframes represent the starting and ending of
the simulation. Again at ‘Animation Manager’, in the Translation:Z field of each
keyframe, elevation values were set according to the water levels of the different
rainfall scenarios. The simulation was run (simulation time was 48 h for Scenario-1
and 24 h for rest of the scenarios), and scenarios were exported in JPEG format.
3.6 Vulnerability and Risk Assessment
Under different hypothetical precipitation scenarios, the risk potential in Khatunganj

is simulated using GIS. Vulnerability can be defined as the degree to which a system
is susceptible to, or unable to cope with, unfavorable climate change impacts, such
as climate variability and extremes. It is determined by the character, magnitude, and
rate of climate change to which a system is subjected to, as well as the sensitivity
and adaptive capacity of that system (IPCC 2001).
Four criteria were chosen to determine the vulnerability of Khatunganj to flooding
caused by different rainfall scenarios. The four criteria are plinth level of the structure,
proximity to Chaktai canal, structure type (kacha, pucca, or semi-pucca), and land
use pattern.
Table 1 shows the respective weightages under each criterion of vulnerability.
Normalization of input data of each criterion is done using the following equation:
X − min(X )
X norm = (3)
max(X ) − min(X )
Here, X is the criterion of vulnerability. The equation used to calculate vulnera-

bility index for each structure was:
Table 1 Weightages for each criterion of vulnerability

Plinth w Proximity w Structure Type w Land Use w
3 ft (1 m) below 6 1.05 m< 1 Paved (pucca) 1 Residential 1
2 ft (0.6 m) below 5 0.70 m–1.05 m 2 Semi-paved 2 Community 1
(semi-pucca) Service
1 ft (0.3 m) below 5 0.35 m–0.70 3 Tin shed 3 Mixed Use 2
(kacha)
Road level 4 0 m–0.35 4 Service Activity 2
1 ft (0.3 m) above 3 Manufacturing 3
2 ft (0.6 m) above 2 and
Commercial
activity
3 ft (1 m) above 1 Commercial 3
activity
n
Wi X i
Vulnerability = i=1
(4)
n
Here, X is the criterion of vulnerability, W is the weight of individual criterion,
and n is the number of criteria. The equation has been adopted and modified from the
Household Sector Approach of Leon and Carlos for this research (Léon and Carlos
2006).
Vulnerability = (Plinth + Proximity + 0.8* Land Use + 0.8* Structure Type)/4

(5)
Weights of the criteria were given as per expert opinion.

A total of 224 structures were identified at the study site, and for each of the
three flood scenarios of vulnerability analysis, these structures were classified into
four classes of vulnerability: not vulnerable, less vulnerable, medium vulnerable,
and very vulnerable.
For risk profiling, three categories have been documented—low rainfall which is
lower than normal and it is 50 mm per day of hypothetical rain that creates up to
2.8 m water level; average rainfall that causes up to 3.5 m water level; and heavy
rainfall that generates more than 3.5 m water level from the datum (the sea level).
Table 2 represents an overview of datasets used for the study and their collection
sources, and Fig. 6 depicts a methodological flowchart.
Table 2 Data used with its source for the study

Parameter Dataset Data source
Watershed delineation DEM ASTER (USGS)
Drainage network Primary data
(Field visit)
Bathymetry Chaktai cannel cross-section (at 89 CEGIS
nodes)
Cannel Water level Water level of Chaktai outfall of
2017
Cannel Water discharge Discharge of Chaktai outfall for
3 months (June, July and August) of
1983
Precipitation Rainfall data of Chattogram Meteorological department
(1980–2017) (BMD)
Landuse Household GIS dataset Chattogram Development
Authority (CDA)
Fig. 6 Methodological
flowchart
4 Results
4.1 Catchment Delineation
Figure 7 shows the catchment boundary of Muradpur-Chawkbazar basin with its

divides, and Fig. 8 is focused on Khatunganj. 32 sub-basins have been identified in this
Mudarpur-Chawkbzar basin. Since drainage water from these sub-basins contributes
to the Chaktai canal, and also affects the waterlogging scenario at Khatunganj, all
these sub-catchments were considered in the hydrological model.
4.2 Inundation Scenario with Actual and Design Rainfall
Design rainfall, which was generated from IDF curves, is used for every sub-
catchment in the model. Table 3 presents the rainfall values that have been consid-
ered for the modeling in different scenarios and Figs. 9, 10, 11, and 12 depict the
hyetographs of actual and design rainfall.
Fig. 7 Catchment area map adjacent to Chaktai
4.3 Discharge and Water Level
In different scenarios, the discharges (hydrograph) and water levels at the nodes
(node 69 & 70) near Khatunganj have been considered for the study and presented
in Fig. 13. For each scenario, the maximum discharge and WL have been reported.
4.4 Validation of SWMM Model
The model was run for the maximum rainfall as observed in the year 2017, since
water level data is only available for that year. Corresponding inundation pattern
was simulated using ArcGIS. Due to resource constraints and the unavailability
of any other data for calibrating and validating the model, an indirect validation
approach was used using satellite image. The simulated water inundation vicinity
to Khatunganj area was compared with the actual flood inundation extent obtained
from the satellite image of Sentinel-1. Sentinel-1 images have 10 m spatial resolution
and perform well in case of inundation mapping in a cloudy environment (Tanim and
Mullick 2017).
With the knowledge of an actual rainfall event on 05 July 2017, the model was
run, and the extent of inundation was estimated. It was then compared with the
Fig. 8 Sub-catchment area containing Khatunganj
satellite image. Based on satellite image analysis, 32.35% of the study area was
found as inundated on the day, whereas the simulation study found an inundation of
around 39.11% of the area. The obtained results are quite compatible and the model
can be considered as calibrated. Figure 14 presents the comparison between actual
inundation as obtained from the satellite image and the simulated inundation from
the actual rainfall on 05 July in 2017.
Table 3 Rainfall values considered for different scenarios

Scenario SL no Scenario description Rainfall value
S-1 Max rainfall event in 2017 distributed over 24 h (Actual 222 mm
Scenario)
S-1 (A) Max rainfall event in 2017 (2 h, High Tide) 222 mm
S-2 2-h 20 year return period rainfall (High Tide) 200.80 mm
S-4 2-h 50 year return period rainfall (Low Tide) 244.98 mm
S-5 2-h 50 year return period rainfall (50% rainfall decrease and 122.49 mm
High Tide)
S-6 2-h 50 year return period rainfall (50% rainfall decrease and 122.49 mm
Low Tide)
S-8 2-h 50 year return period rainfall (considering Climate change 271.92 mm
11% rainfall increment)
Fig. 9 Hyetographs of scenario 1
4.5 Inundation Maps
Percentages of inundated area are presented in Fig. 15. At hourly maximum rainfall,
counted according to the actual scenario of 2017 (S-1), 39.1% area of Khatunganj got
inundated. A 2-h, 50-year return period rainfall, considering 50% rainfall decrease
and low tide (S-6), inundated 24.6% area of Khatunganj. At an extreme case, for 2-h
50-year return period rainfall, considering climate change and 11% rainfall increment
(S-8), 49.08% of the total area of Khatunganj was found to be submerged. At the
end, an inundation scenario (S-10) was assumed, which inundated 12.18% area of
Khatunganj with a flood level of 2.56 m.
Fig. 11 Hyetographs of scenario 3 & 4
4.6 Vulnerability and Risk Analysis
Figures 16 represent the structural classification based on the four criteria i.e., plinth,
proximity, structure type, and land use.
Figures 17, 18, and 19 demonstrate the vulnerability for different categories of
rainfall. Table 4 documents the number of structures at risk due to different flood
scenarios. In case of low rainfall, 122 structures were found as not vulnerable, and
27 structures to be very vulnerable. Besides, there were 33 less vulnerable and 42
medium vulnerable structures. In the event of average rainfall, 37 structures were
found as not vulnerable, and 83 structures were less vulnerable. The number of
medium and very vulnerable structures were 74 and 30, respectively. At the time
of heavy rainfall, every structure was vulnerable to some extent. There were 68
less vulnerable structures and 85 medium vulnerable structures. Additionally, 71
structures of Khatunganj were identified as very vulnerable at the time of heavy
rainfall.
5 Discussion
Analysis shows that in all the different scenarios, the water level exceeds 3 m from the
sea level (the datum), whereas the existing road level in Khatunganj varies between
2.9 m and 3.05 m from the datum. So, in the event of average to heavy rainfall, the
Chaktai canal becomes overtopped, and the adjacent areas get flooded. The structures
on both side of Khatunganj road have 3 types of plinth levels, namely, plinth elevation
near to road level, below road level, and above road level. In fact, the site is very
much historic, and majority of the structures in the area are very old. On the other
hand, the road level is upgraded by the city authority from time to time. This results
in a number of structures being lower than the road level. However, the plinth level
of new structures is a bit high, even going up to 3 feet higher than the existing road
level.
It was observed that tide events had a vital influence on the inundation for the
same extent of rainfall. A 2-h 50-year return period rainfall creates a flood level
3.94 m at the time of high tide (S-3) and inundates 46.14% of Khatunganj area. On
the contrary, the same amount of rainfall causes a 3.5 m flood at the period of low tide
(S-4), submerging 36.23% of the total area. Similar influence was observed in case
of scenarios S-5 and S-6. Considering 50% rainfall decrease, a 2-h 50-year return
period rainfall generates 3.84 m flood at the time of high tide (S-5), but the same
Scenario S-1: Max rainfall event in 2017 (Actual Scenario)
Maximum Q = 179.48 CMS Maximum WL at Node 69 = 3.62 m

Scenario S-1(A): Max rainfall event in 2017 (2 hour, High Tide)

Scenario S-2: 2-hour 20 year return period rainfall (High Tide)

Fig. 13 Discharge and Water level for different scenario analyzed

Scenario S-4: 2-hour 50 year return period rainfall (Low Tide)
Maximum Q = 955.52 CMS Maximum WL at Node 69: 3.5 m

Scenario S-5: 2-hour 50 year return period rainfall (considering 50% rainfall decrease) (High Tide)

Scenario S-6: 2-hour 50 year return period rainfall (considering 50% rainfall decrease) (Low Tide)

Fig. 13 (continued)
Scenario S-8: 2-hour 50 year return period rainfall (considering Climate change 11% rainfall
increment)
Fig. 13 (continued)
Fig. 14 Comparison between a Actual Flood Scenario obtained from Satellite and b Simulated
Flood Scenario
rainfall generates a flood level of 3.09 m during low tide (S-6). Inundated area at the
span of high tide (S-5) is 44.08%, and it drops down to 24.6% at the time of low tide.
Majority of the structures in Khatunganj area fall in the medium vulnerable class
for all three rainfall cases. 122 structures out of 224 i.e., 54.4% of all structures in the
area are not vulnerable at the time of low rainfall. However, event of high rainfall,
all the structures have been found as vulnerable to various extent. Only 27 structures
are very vulnerable in case of low rainfall. The number of very vulnerable structures
rose up to 30 at the time of average rainfall, whereas 71 structures i.e., 31.7% of all
structures in the area are very vulnerable at the time of heavy rainfall.
Fig. 15 Inundation pattern as obtained from model for different scenarios

Fig. 15 (continued)
6 Conclusion
This study did an extensive analysis of the recurrent waterlogging problem in Khatun-
ganj commercial area. A small rainfall results in an inundation in the area and
the consequences are substantial economic loss in the commercial area. The study
enumerated the vulnerability of the area due to waterlogging. To carry out such
analysis, a hydrologic-hydraulic model was first developed to provide an inunda-
tion scenario and from the inundation scenario, vulnerability of each structure was
assessed.
A model was formulated on the SWMM model platform, which provided
discharge and water level at different canal points. Nine different scenarios on
different rainfall events were simulated, where an actual rainfall event was taken
from 05 July 2017, and remaining scenarios were set up as design rainfall events.
Due to data scarcity, the model was calibrated using a satellite image of inundation
extent from the actual rainfall.
In the event of any average to heavy rainfall, the Chaktai canal becomes over-
topped, resulting in flooding of the surrounding areas. Average rainfall generates a
WL of 3–3.5 m and the road level is about 2.9 to 3.1 m. Therefore, such a WL results
in inundation of the Khatunganj main road with a water depth of 1 to 15 inch. Heavy
rainfall with climate change scenario gives around 4.09 m WL, indicating around
1 m depth of water inundation of the road. All structures on both sides of the road
will be submerged in such a scenario.
Majority of the structures in the area can be categorized as medium vulnerable for
all three rainfall cases. In case of heavy rainfall, all structures are vulnerable to some
extent. Water retention options, increased canal capacity, and regular maintenance
of canal are extremely important for managing such a crisis.
Fig. 16 Area Classification based on a plinth of the structure, b proximity to Chaktai canal, c
structure type, d land use
Fig. 17 Vulnerability for low rainfall
Fig. 18 Vulnerability for average rainfall

Fig. 19 Vulnerability for heavy rainfall
Table 4 Number of
Low rainfall Average Heavy rainfall
structures at risk due to
rainfall
different flood scenarios
Not 122 37 0
vulnerable
Less 33 83 68
vulnerable
Medium 42 74 85
vulnerable
Very 27 30 71
vulnerable
Total 224 224 224
References
Alexander, Kane, Suresh Hettiarachchi, Yixiao Ou, and Ashish Sharma. 2019. Can Integrated Green
Spaces and Storage Facilities Absorb the Increased Risk of Flooding Due to Climate Change in
Developed Urban Environments? Journal of Hydrology 579. Elsevier B.V.: 124201. https://doi.
org/10.1016/j.jhydrol.2019.124201.
Barua, Dwaipayan. 2021. Khatunganj Lost over Tk 514cr to Waterlogging in 2020: Study. The
Daily Star.
Berndtsson, R., P. Becker, A. Persson, H. Aspegren, S. Haghighatafshar, K. Jönsson, R. Larsson,
et al. 2019. Drivers of Changing Urban Flood Risk: A Framework for Action. Journal of Envi-
ronmental Management 240 (October 2018). Elsevier: 47–56. https://doi.org/10.1016/j.jenvman.
2019.03.094.
Burrough, Peter A., Rachael A. McDonnell, and Christopher D. Lloyd. 1998. Principles of
Geographical Information Systems. Oxford University Press. https://books.google.com.bd/
books?hl=en&lr=&id=kvoJCAAAQBAJ&oi=fnd&pg=PP1&dq=Burrough,+P.+A.,+%26+McD
onnell,+R.+A.+(1998).+Principles+of+geographical+information+systems.+Oxford:+Oxford+
University+Press&ots=aKI63ms3HG&sig=cxKiRofsF8-2ctO1IChRca9V6EM&redir.
Chowdhury, Mitoon. 2011. SPECIAL Century-Old Khatunganj Loses Its Glory. Bdnews24.Com.
Deng, Yu, Wei Qi, Bojie Fu, and Kevin Wang. 2020. Geographical Transformations of Urban Sprawl:
Exploring the Spatial Heterogeneity Across Cities in China 1992–2015. Cities 105 (January).
Elsevier: 102415. https://doi.org/10.1016/j.cities.2019.102415.
Dutta, Dushmanta, Srikantha Herath, and Katumi Musiake. 2003. A Mathematical Model for Flood
Loss Estimation. Journal of Hydrology 277 (1–2): 24–49. https://doi.org/10.1016/S0022-169
4(03)00084-2.
Fleming, K.M., P. Tregoning, M. Kuhn, A. Purcell, and H. McQueen. 2012. The Effect of Melting
Land-Based Ice Masses on Sea-Level Around the Australian Coastline. Australian Journal
of Earth Sciences 59 (4). Taylor & Francis: 457–467. https://doi.org/10.1080/08120099.2012.
664828.
Han, Su Qin, Yi Yang Xie, Da Ming Li, Pei Yan Li, and Mei Ling Sun. 2006. Risk Analysis and
Management of Urban Rainstorm Water Logging in Tianjin. Journal of Hydrodynamics 18 (5).
China Ocean Press: 552–558. https://doi.org/10.1016/S1001-6058(06)60134-0.
Hasan, Zaheed, Sabiha Akhter, and Mamunul Islam. 2014. Climate Change and Trend of Rainfall
in the South-East Part of Coastal Bangladesh. European Scientific Journal 10 (2): 25–39.
https://www.researchgate.net/publication/260002503_CLIMATE_CHANGE_AND_TREND_
OF_RAINFALL_IN_THE_SOUTH-EAST_PART_OF_COASTAL_BANGLADESH.
Hirabayashi, Yukiko, Shinjiro Kanae, Seita Emori, Taikan Oki, and Masahide Kimoto. 2010. Global
Projections of Changing Risks of Floods and Droughts in a Changing Climate. Hydrological
Sciences Journal 53 (4). Taylor & Francis Group: 754–772. https://doi.org/10.1623/HYSJ.53.
4.754.
IPCC. 2001. Climate Change 2007: Impacts, Adaptation and Vulnerability.
Jang, Suhyung, Minock Cho, Jaeyoung Yoon, Yongnam Yoon, Sangdan Kim, Geonha Kim,
Leehyung Kim, and Hafzullah Aksoy. 2007. Using SWMM as a Tool for Hydrologic Impact
Assessment. Desalination 212 (1–3): 344–356. https://doi.org/10.1016/j.desal.2007.05.005.
Jenson, S.K., and J.O. Domingue. 1988. Extracting Topographic Structure from Digital Elevation
Data for Geographic Information System Analysis, Vol. 54.
Kabir, Ekram. 2016. What’s so Special About Chittagong. Dhaka Tribune.
Kabir, Shah Md Imran, and Mostafa Ali. 2017. Hydrodynamic and Morphological Simulation
of Karnafuli River Using Cche2d Model. In 6th International Conference on Water & Flood
Management (ICWFM-2017). https://www.researchgate.net/publication/328007509_HYDROD
YNAMIC_AND_MORPHOLOGICAL_SIMULATION_OF_KARNAFULI_RIVER_USING_
CCHE2D_MODEL.
Kleist, L., A.H. Thieken, P. Köhler, M. Müller, I. Seifert, D. Borst, and U. Werner. 2006. Estimation
of the Regional Stock of Residential Buildings as a Basis for a Comparative Risk Assessment in
Germany. Natural Hazards and Earth System Sciences 6 (4): 541–552. https://doi.org/10.5194/
nhess-6-541-2006.
Léon, Villagrán de, and Juan Carlos. 2006. Vulnerability: A Conceptional and Methodological
Review - UNU Collections. Bonn: UNU- EHS. http://collections.unu.edu/view/unu:1871.
Miller, James D., Hyeonjun Kim, Thomas R. Kjeldsen, John Packman, Stephen Grebby, and Rachel
Dearden. 2014. Assessing the Impact of Urbanization on Storm Runoff in a Peri-Urban Catchment
Using Historical Change in Impervious Cover. Journal of Hydrology 515. Elsevier B.V.: 59–70.
https://doi.org/10.1016/j.jhydrol.2014.04.011.
Ministry of Foreign Affairs of the Netherlands. 2019. Climate Change Profile Bangladesh. https://
doi.org/10.1080/17565529.2014.977761.
Mishra, Surendra Kumar, and Vijay P. Singh. 2013. Soil Conservation Service Curve Number
(SCS-CN) Methodology. Water Science & Technology Library 42.
Nasher, N.M. Refat, and M.N. Uddin. 2015. Maximum and Minimum Temperature Trends Variation
over Northern and Southern Part of Bangladesh. Journal of Environmental Science and Natural
Resources 6 (2). Bangladesh Journals Online (JOL): 83–88. https://doi.org/10.3329/jesnr.v6i2.

22101.
Nayeb Yazdi, Mohammad, Mehdi Ketabchy, David J. Sample, Durelle Scott, and Hehuan Liao.
2019. An Evaluation of HSPF and SWMM for Simulating Streamflow Regimes in an Urban
Watershed. Environmental Modelling and Software 118 (January). Elsevier: 211–225. https://
doi.org/10.1016/j.envsoft.2019.05.008.
Oliveri, Elisa, and Mario Santoro. 2000. Estimation of Urban Structural Flood Damages: The Case
Study of Palermo. Urban Water 2 (3): 223–234. https://doi.org/10.1016/s1462-0758(00)00062-5.
Peng-zhu, X.U., Jiang Tong, and King Lorenz. 2000. Hydrologic/Hydraulic Modelling and Flood
Risk Analysis for the West Tiaoxi. Chinese Geographical Science 10 (4): 309–318.
Quan, Rui Song. 2014. Rainstorm Waterlogging Risk Assessment in Central Urban Area of Shanghai
Based on Multiple Scenario Simulation. Natural Hazards 73 (3): 1569–1585. https://doi.org/10.
1007/s11069-014-1156-x.
Quddusi, Kazi S.M. Khasrul Alam. 2017. Water-Logging in Chittagong: Glimpses of Losses. Daily
Sun.
Schwab, G.O., D.D. Fangmeier, W.J. Elliot, and R.K. Freveret. 1993. Soil and Water Conservation
Engineering. New York: Wiley.
Shi, Yong, Chun Shi, Shi Yuan Xu, A. Li Sun, and Jun Wang. 2010. Exposure Assessment of
Rainstorm Waterlogging on Old-Style Residences in Shanghai Based on Scenario Simulation.
Natural Hazards 53 (2): 259–272. https://doi.org/10.1007/s11069-009-9428-6.
Tanim, Ahad Hasan, and Md. Reaz Akter Mullick. 2017. Application of Synthetic Aperture Radar
Imagery for Inundation Mapping : A Case of Sylhet Flood Event 2017, no. September.
Wanniarachchi, S.S., and N.T.S. Wijesekera. 2012. Using SWMM as a Tool for Floodplain Manage-
ment in Ungauged Urban Watershed. Engineer: Journal of the Institution of Engineers, Sri Lanka
45 (1): 1. https://doi.org/10.4038/engineer.v45i1.6944.
Xue, Fengchang, Minmin Huang, Wei Wang, and Lin Zou. 2016. Numerical Simulation of Urban
Waterlogging Based on FloodArea Model. Advances in Meteorology 2016. https://doi.org/10.
1155/2016/3940707.
Yin, Zhane, Jie Yin, Shiyuan Xu, and Jiahong Wen. 2011. Community-Based Scenario Modelling
and Disaster Risk Assessment of Urban Rainstorm Waterlogging. Journal of Geographical
Sciences 21 (2). Springer: 274–284. https://doi.org/10.1007/s11442-011-0844-7.
Hydrometeorological Hazards and Risk
A Remote Sensing-Based Approach
for Analysis of Dry and Wet Periods
of Bangladesh Based on Standardized
Precipitation Index During 1981–2020
Saumik Mallik
Abstract Bangladesh, having a subtropical monsoon climate with wide seasonal

fluctuations in rainfall, is susceptible to recurrent dry and wet periods. This study
aims to measure the spatial and temporal variability of dry/wet periods at district
level, based on timescale and intensity by analyzing satellite data from CHIRPS
precipitation record for the past 40 years to calculate the standardized precipitation
index (SPI). The frequencies of moderate, severe, and extreme dry and wet periods
for both short and long timescales for individual districts are calculated by extracting
and analyzing the individual SPI time series of the 64 districts in GIS and R through
geospatial analysis and verified using past records of well-defined drought and flood.
The results demonstrated that, within a period of 100 years, districts from Khulna
and Rangpur division can anticipate about 18 droughts and 17 floods, respectively, of
different intensities lasting 12 months on average, which is the highest in Bangladesh.
As for droughts and floods lasting 3 months, districts from Barisal and Dhaka divi-
sion lead, respectively, with about 66 droughts and 67 floods expected on average.
The findings demonstrate that, based on timescale and intensity, the occurrences of
dry/wet periods varied to a great extent across Bangladesh during the past 40 years.
Keywords Bangladesh · Remote sensing · Standardized precipitation index ·

Droughts and floods · Extreme events
1 Introduction
Global climate change, encompassing steady as well as abrupt shifts, has a consid-
erable influence on land surface-atmosphere dynamics and local socio-economic
development. Global warming, or progressive climate change, has a major impact
on the hydrological cycle and atmospheric circulatory systems, altering the intensity
and geographic distribution of precipitation (Parry et al. 2007; Arnell 1999), that in
turn disrupts the local meteorological conditions and the regional agriculture sector.
S. Mallik (B)
Department of Civil Engineering, Bangladesh University of Engineering and Technology, Dhaka
1000, Bangladesh
124 S. Mallik
Additionally, as a result of climate change, the frequency and intensity of extreme

weather occurrences may alter (Rosenzweig et al. 2001), resulting in a variety of
meteorological disasters such as floods, droughts, and rainstorms.
Despite considerable scientific and technological advancements and improve-
ments in environmental management over the twentieth century, mankind continues
to be affected by the repercussions of these meteorological disasters on a global scale
(Krysanova, Vetter, and Hattermann 2008). Climate change is only one of several
indicators of our previous inability to accomplish sustainable development (Green
2010), other indicators include the approach of peak oil (IEA 2009) and peak phos-
phorus (Cordell et al. 2009), as well as the challenge of satisfying global food needs
(Molden 2013). It is essential to analyze the spatial and temporal variability of local
extreme hydroclimatic events in order to garner a better understanding of recent
climatic variations, their appearance in local areas, and to regularly keep track of
drought and flood events. This insight will significantly assist in the advancement of
integrated water resource management.
Conventional drought monitoring entails the study of rain-gage network measure-
ments, which are more reliable when coupled with long-term satellite records (Khan
et al. 2014; Bechle et al. 2013). Precipitation data from several meteorological stations
in Bangladesh, provided by the Bangladesh Meteorological Department (BMD),
is often utilized in a variety of research. However, this technique has numerous
drawbacks, including spatial contiguity, data inaccessibility in remote regions, high
maintenance costs, and near-real-time data unavailability to ordinary users (Pandey
et al. 2020). Alternatively, spatially contiguous precipitation data enables a more
accurate evaluation of drought and flood occurrences. Moreover, high-resolution
satellite data has been frequently employed in recent years to characterize drought
episodes spatiotemporally (Tao et al. 2016; Gao et al. 2018; Shrestha et al. 2017).
The Climate Hazards Group’s Infrared Precipitation with Station (CHIRPS) data is a
newly developed high-resolution precipitation product derived from satellite data and
integrated with weather stations observations. It has demonstrated significant poten-
tial in a variety of hydrological studies, including drought monitoring (Funk et al.
2014, 2015). Numerous regions in Asia, specifically in India, China, and Nepal,
are demonstrating the use of CHIRPS records for real-time retrospective drought
and flood study. Shrestha et al. (2017) used CHIRPS precipitation data to monitor
drought in Nepal’s Koshi River watershed and suggested the data’s usefulness for
such research. Additionally, the CHIRPS-based SPI is shown to have the capability
of identifying drought events and their characteristics, suggesting it can be used as a
novel source of information for drought monitoring across diverse regions of China
(Gao et al. 2018; Zhong et al. 2019; W. Wu et al. 2019). Pandey et al. (2020) assessed
the performance of high-resolution satellite precipitation datasets, including TRMM,
PERSIANNCDR, and CHIRPS, in order to determine their effectiveness for drought
monitoring. It was found that CHIRPS data had the highest accuracy when compared
to the gridded gage dataset over their study area of central India.
Many drought-monitoring indicators have been developed (Ahmed et al. 2019;
Alamgir et al. 2015). The standardized precipitation index (SPI) is the most straight-
forward, statistically robust, comprehensible, and independent of meteorological
A Remote Sensing-Based … 125
variables among these (Keyantash and Dracup 2002). SPI is a probability-based

statistic that indicates how much a given period’s accumulative precipitation deviates
from the average state. The index is spatially independent and performs excellently
for expressing precipitation anomalies (McKee et al. 1993). In comparison to other
physical processes-based indices and methodologies, such as the palmer drought
severity index (PDSI), the SPI is straightforward to compute and implement. It only
uses precipitation as input data, which eliminates the need for parameter calibration,
making it especially well-suited for drought/flood monitoring in locations with a
scarcity of hydrological data (Yuan and Zhou 2004). In a multitude of countries and
regions, the SPI has been extensively utilized to classify dry and wet conditions,
including Canada (Quiring and Papakryiakou 2003), China (Du et al. 2013), Iran
(Moradi et al. 2011; Nafarzadegan et al. 2012), Italy (Piccarreta et al. 2004; Vergni
and Todisco 2011), Korea (Min et al. 2003; Kim et al. 2009), and the United States
(H. Wu et al. 2007).
Bangladesh’s rainfall pattern is shifting as a result of global climate change
(Shahid and Behrawan 2008; Rimi et al. 2019). This has resulted in an increase in
the country’s extreme weather events (Basher et al. 2018) and hydrological disasters
(Mohsenipour et al. 2018). As a result of changes in rainfall patterns, Mohsenipour
et al. (2018) noticed a drop in the return period of droughts from 1985 to 2014
at stations where it was high from 1961 to 1990, and vice versa in Bangladesh’s
most drought-prone areas. Bangladesh, which is located in tropical South Asia, has
suffered many catastrophic droughts and floods as a result of precipitation shortages
and excesses (Shahid and Behrawan 2008). These extreme occurrences resulted in
protracted water scarcity or inundation, affecting agriculture, the environment, and
public health (Keyantash and Dracup 2002). Population increases, agricultural expan-
sion, land use changes, and industrial development all amplify the impact of these
disasters (“Glossary of Terms” 2014). This necessitates a thorough knowledge of
the geographical and temporal diversity associated with dry/wet cycles on various
timescales.
Even though a substantial body of research has been conducted in Bangladesh
on the spatial and temporal characteristics of droughts and floods (Mohsenipour
et al. 2018; Rimi et al. 2019), no studies have been performed on a national scale in
Bangladesh that quantify the spatial and temporal variability of dry/wet periods on a
districts basis and focus on both timescale and intensity utilizing satellite data. The
novelty of this work is that it uses a high-resolution precipitation product derived from
satellite data in conjunction with rain-gage observations from 1981 to 2020 to explore
the frequency of both dry and wet periods of varying intensities over both short-
and long-term timescales using SPI. Remote sensing techniques have established
themselves as an increasingly efficient strategy for researchers and policymakers
(Kafy et al. 2021), and this study will contribute to the development and maintenance
of the country’s water resources, as well as limit the impact of extreme events to
ensure sustainable management and planning.
126 S. Mallik
2 Materials and Method
2.1 Study Area
The study area encompasses Bangladesh, situated in South Asia between latitudes
20°34 and 26°38 N and longitudes 88°01 and 92°41 E. Bangladesh is bounded
by India on the west, north, and northeast and bordered on the south by Myanmar.
The Bay of Bengal, with its extensive coastline, serves as the southern boundary.
Excluding the higher elevations at the far southeast and the Sylhet division in the
northeast, the majority of Bangladesh is less than 10 m above sea level. Agricultural
land prevails in the country, since more than 50% of the entire geographical land was
estimated to be cropped areas during 2019–2020 (Bangladesh Bureau of Statistics
2021). Bangladesh is particularly vulnerable to global climate change’s escalating
consequences. It is often struck by natural calamities, including flooding, drought,
cyclones, and tidal bores (Bangladesh Bureau of Statistics 2015). Bangladesh has
recently suffered drought on a frequent basis; on average, Bangladesh faces drought
once every 2.5 years (Mondol et al. 2017).
2.2 Datasets
The preceding four decades (1981 to 2020) were chosen as study period for this
work. This study makes use of gridded precipitation data from The Climate Hazards
Group’s Infrared Precipitation with Stations (CHIRPS) dataset. Estimates obtained
from satellite data produce area-averaged values that typically underestimate the
intensity of extreme precipitation events owing to complex geography. Precipita-
tion grids generated from station data, on the other hand, suffer in more remote
regions with fewer rain-gage stations. To construct gridded rainfall time series for
seasonal drought monitoring and trend analysis, CHIRPS data contains in-house
climatology—CHPclim, 0.05° resolution satellite images, and in situ station data
from 1981 to the near-present. Its algorithm is based on a 0.05° climatology that
(i) incorporates satellite data to represent sparsely gaged locations, (ii) incorporates
daily, pentadal, and monthly 0.05° CCD-based precipitation estimates from 1981 to
the present, (iii) blends station data to generate a preliminary information product
with an average latency of about 2 days and a final product with an average latency
of about 3 weeks, and (iv) uses an unique blending procedure to assign interpolation
weight utilizing the spatial correlation structure of CCD-estimates (Funk et al. 2015).
2.3 The Standardized Precipitation Index (SPI)
Drought is quantified on the basis of the standardized precipitation index (McKee

et al. 1993), which has also been recommended by the WMO for drought moni-
toring (WMO 2012). Additionally, wetter and drier climates can be portrayed in the
same way since the SPI is standardized; hence, utilizing the SPI, wet periods can be
analyzed and monitored just as well. The SPI transformation begins by applying a
backward-looking moving average with a window length τ to the monthly precipita-
tion data, resulting in accumulated precipitation time series. The SPI was computed
for each month of this sequence of years at timeframes of τ = 3 and τ = 12 months
in this study. The rest of this paper will refer to index values derived for window
length using the notation SPI- τ. To determine SPI-3 and SPI-12, a new series was
constructed for each month of the calendar year with components equal to the corre-
sponding precipitation moving sums for each individual pixel within the study area.
For example, the 3-month SPI for February 2000 was computed using the total
precipitation from November 1999 to February 2000. Similarly, the 12-month SPI
for February 2000 was calculated using the total precipitation from March 1999 to
February 2000.
The cumulative precipitation series is then normalized individually for each
calendar month. This is accomplished by assembling data from a single calendar
month in a base period (e.g., all February values from 1981 to 2020) and fitting the
data to a gamma distribution. The parameters of the gamma distribution are used
to transform all the values of the associated months to those of the standard normal
distribution, resulting in a mean SPI of zero for the pixel and desired period (Stagge
et al. 2015; Edwards 1997). The area-averaged SPI time series for each of the 64
districts is then calculated in QGIS using the pixels contained inside the district
borders.
The lowest bound for precipitation is zero, and their empirical distributions exhibit
a positive skewness. As a result, the gamma distribution has been a popular choice
for calculating the SPI. The gamma distribution was found to fit long-term precipi-
tation data well, in accordance with earlier research (Stagge et al. 2015; Livada and
Assimakopoulos 2007; Thom 1958). The gamma distribution is defined as follows
in terms of a frequency or probability density function:
1
g(x) = x α−1 e−x/β , for x > 0 (1)
β α (α)
where Γ (α) is the gamma function; x is the precipitation; and α and β are the
shape and scale parameters, respectively.
For the 3- and 12-month timescales, the number of months Ni in each class of
drought and flood intensity as defined in Table 1 was calculated. The number of
droughts and floods every hundred years was then estimated as follows:
Ni
Ni,100 = .100 (2)
i, n
128 S. Mallik
Table 1 Meteorological
SPI values Class Probability
drought and flood risks
classification using SPI 2.0 and more Extremely wet 0.023
values (McKee et al. 1993) 1.5 to 1.99 Very wet 0.044
1.0 to 1.49 Moderately Wet 0.092
0.99 to 0.99 Near normal 0.682
−1.0 to −1.49 Moderately dry 0.092
−1.5 to −1.99 Severely dry 0.044
−2 and less Extremely dry 0.023
where N i ,100 = the number of droughts/floods for a timescale i in 100 years.

N i = the number of months with droughts/floods for a timescale i in the n-year
set.
i = timescale (3, 12 months).
n = the number of years in the data set (=40).
Pre-processing of the CHIRPS satellite data included several intermediate steps
such as cropping the study area, merging all netCDF format files into one netCDF
file, editing variable names and units, and were done using BASH script. SPI values
were calculated for each pixel within the study area using the open-source soft-
ware package Climate Indices for Python (Adams 2017), which includes Python
implementations of various climate index algorithms that provide a geographical
and temporal portrait of the intensity of precipitation and temperature irregularities
effective for climate monitoring and research. In QGIS, the pixels within a district
border are used to calculate the area-averaged SPI value for each of the 64 districts.
The resultant dataset is then imported into R, where the frequency of drought/flood
occurrences of various intensities is analyzed. Timescale and intensity-based maps
of long and short-term dry/wet periods are generated from the analysis.
McKee et al. (1993) defined drought intensities resulting from the SPI using the
categorization system shown in the SPI value table below (Table 1). Additionally, they
specified the parameters for a drought occurrence across all timescales. A drought
episode occurs when the SPI remains persistently negative and maintains a magnitude
of −1.0 or less. When the SPI turns positive, the event terminates. Similarly, a
persistent positive SPI value of +1.0 or more is considered a flood episode.
In this study, a 3-month SPI is used to identify short-term or seasonal dry/wet
events, while a 12-month SPI is used to detect long-term dry/wet events. Stream-
flow, reservoirs, and groundwater respond to longer-term precipitation anomalies of
the order of 6 months to 24 months or more, whereas meteorological and soil mois-
ture conditions (agricultural) respond relatively quickly to precipitation anomalies,
typically within 1–6 months. SPIs with a shorter time horizon, such as 3-month SPIs,
can give early warning of drought/flood and aid in determining the severity of the
event.
3 Results
Three frequency-based maps for each of the intensities (moderate, severe, and
extreme) are created for both long- and short-term dry and wet events, and a further
four maps are generated based on the total number of events encountered by each
district regardless of intensities using SPI analysis of 3- (short-term) and 12- (long-
term) months, thus identifying districts with a greater prevalence of drought and
flooding during a 40-year period (1981–2020). The SPI results are such classified into
short- and long-term analysis, as well as three distinct intensities, in order to provide
comparative analysis and to propose appropriate short- and long-term actions.
3.1 Spatial Distribution of the Frequency of Long-Term

Dry/Wet Events in Bangladesh
The SPI-12 results are classed as long-term analysis, illustrating drought and flood
rates for moderate, severe, and extreme intensities over all districts, with red color
symbology emphasizing the most frequently affected areas (Fig. 1). The frequencies
of long-term droughts and floods of individual districts, as well as the averaged
frequencies for each of the seven divisions, are shown in Table 2.
Long-term SPI is a term that refers to the long-term effects of droughts on stream
flow, reservoir storage, and groundwater level, and hence is connected to hydrological
droughts. The incidence of moderate dry and wet periods is higher than the occurrence
of severe or extreme dry/wet periods in the majority of the country’s districts.
The north-eastern district of Sylhet is one of the least affected by long-term
droughts. However, it is observed that its neighboring districts to the west have a
higher frequency of extreme dry occurrences but are less likely to have moderate-
to-severe events. Districts in the north-western Rajshahi and south-western Khulna
regions are more likely to have long-term severe dry episodes, whereas districts
throughout southern Bangladesh, including the Chittagong hill tracts region, have
had regular moderate dry spells throughout the study period. Long-term study reveals
a rise in frequency percentages, as seen in Table 2, which compares distinct drought
and flood categories and their respective frequency percentages for 12 months SPI.
It is found that in places prone to severe to extreme droughts, there is a significant
likelihood of hydrological drought and depletion of ground water and stream flow,
and thus necessitating the adoption of water conservation and management practices
in those areas.
On the other hand, the central region of the country surrounding the capital Dhaka,
extending up to the north-eastern Sylhet region, is more likely to encounter long-term
extreme wet periods. As evidenced by the greater average frequency of these two
divisions in Table 2, the south-west districts of Khulna and the northernmost districts
of Rangpur division are more prone to long-term moderate-to-severe wet events than
other regions.
130 S. Mallik
Fig. 1 Spatial distribution of the frequency (in %) of long-term dry periods (a) and frequency of
long-term wet periods (b) of different intensities (extreme, severe, and moderate)
3.2 Spatial Distribution of the Frequency of Short-Term

Dry/Wet Events in Bangladesh
The results of the 3-month SPI are classified as short-term analysis (Fig. 2) since
they primarily give estimates of the effects of dry/wet periods on a short timescale.
Short-term drought and flood frequencies for each individual districts and divisions
were calculated for moderate, severe, and extreme intensities, as shown in Table 3,
and districts having a greater drought/flood frequency were highlighted in red on the
maps. The 3-month SPI is useful for estimating short- and medium-term moisture
conditions in agriculture, since it offers an estimate of seasonal precipitation, which
indicates a deviation from total precipitation. For instance, at the end of February,
the 3-month SPI compares the December–January–February precipitation total for
Table 2 The frequency of occurrences of long-term dry and wet periods of various intensities in
different districts in Bangladesh between the years 1981 and 2020
Division District Extreme Severe Moderate Extreme Severe Moderate
dry (%) dry (%) dry (%) wet (%) wet (%) wet (%)
Barisal Barguna 0.83 6.25 8.13 1.88 4.17 10
Barisal 1.67 2.29 12.29 3.33 5.21 6.04
Bhola 1.88 2.71 10.42 2.29 5.63 7.08
Jhalokati 1.88 3.13 10.63 2.5 5.63 9.38
Patuakhali 0.63 5.83 7.71 2.08 5.83 8.75
Pirojpur 1.67 3.33 12.08 1.04 6.88 8.33
Average 1.43 3.92 10.21 2.19 5.56 8.26
Chittagong Bandarban 0.21 3.33 10.42 2.29 5.83 6.67
Brahamanbaria 1.46 3.75 6.25 3.13 4.17 7.92
Chandpur 1.67 1.25 9.38 3.96 2.29 11.25
Chittagong 0.21 4.17 9.58 2.08 4.38 8.13
Cumilla 1.25 1.88 11.46 3.33 3.13 9.58
Coxs Bazar 0.21 2.29 7.29 3.75 2.29 5.42
Feni 1.46 3.54 7.71 2.5 4.58 9.58
Khagrachhari 1.25 1.67 11.88 2.71 4.58 8.13
Lakshmipur 1.46 2.5 9.79 3.54 4.38 7.92
Noakhali 1.46 2.71 10.42 2.29 4.17 10
Rangamati 0.42 2.92 11.25 2.08 3.96 8.13
Average 1.01 2.73 9.58 2.88 3.98 8.43
Dhaka Dhaka 1.46 2.29 8.13 4.58 2.29 6.04
Faridpur 1.67 2.71 9.79 2.92 3.13 9.58
Gazipur 1.88 2.29 8.33 4.38 1.67 7.5
Gopalganj 1.88 2.71 11.46 3.13 3.54 9.79
Jamalpur 1.67 5.63 7.29 1.25 5 7.92
Kishoreganj 2.5 2.5 5.83 3.33 3.33 6.25
Madaripur 1.88 3.13 8.54 3.54 3.75 9.58
Manikganj 1.46 1.67 8.13 4.38 2.08 8.54
Munshiganj 1.67 2.71 11.25 3.75 2.71 9.17
Mymensingh 2.29 2.29 5 2.92 4.38 6.88
Narayanganj 1.67 3.75 7.08 4.58 2.5 7.5
Narsingdi 2.08 3.75 5.42 3.54 4.38 6.25
Netrakona 3.54 2.08 5.63 1.88 3.75 6.46
Rajbari 0.63 5 8.13 3.33 1.88 10.63
Shariatpur 1.88 1.67 10.63 3.54 3.75 8.75
Sherpur 1.67 4.38 8.33 1.25 5.83 8.33
(continued)
132 S. Mallik
Table 2 (continued)
Tangail 1.67 2.29 8.13 4.17 2.5 7.5
Average 1.85 2.99 8.06 3.32 3.32 8.04
Khulna Bagerhat 0.63 6.88 9.79 2.29 5.83 6.67
Chuadanga 1.88 2.29 11.46 2.29 2.92 8.96
Jessore 2.5 4.17 9.17 0.83 6.04 10.42
Jhenaidah 2.29 2.08 12.08 2.29 4.17 8.75
Khulna 0.63 5.63 12.71 1.04 6.46 7.92
Kushtia 1.88 4.17 12.29 2.71 2.5 7.71
Magura 0.42 5 13.33 2.5 5 10
Meherpur 2.08 4.38 10.83 2.5 2.08 8.33
Narail 1.46 3.75 13.75 1.67 5.21 10
Satkhira 0.63 4.58 14.79 0.63 6.67 8.13
Average 1.44 4.29 12.02 1.88 4.69 8.69
Rajshahi Bogra 0.83 7.5 10 1.46 3.96 8.75
Joypurhat 0.83 7.29 7.92 2.29 5 6.67
Naogaon 1.04 7.29 7.08 2.71 4.58 5.63
Natore 1.04 6.88 8.33 3.13 3.13 7.08
Nawabganj 2.08 7.92 5.63 1.04 6.67 6.04
Pabna 0.21 6.46 8.33 3.33 2.5 7.08
Rajshahi 0.42 8.13 8.33 2.92 4.58 6.25
Sirajganj 0.42 4.17 11.04 3.13 2.92 9.38
Average 0.86 6.96 8.33 2.5 4.17 7.11
Rangpur Dinajpur 2.29 3.96 6.46 2.08 3.54 10.83
Gaibandha 0.83 7.5 7.92 1.25 4.79 11.04
Kurigram 2.29 4.38 10.21 0.83 3.33 9.79
Lalmonirhat 2.29 2.5 9.17 0.63 6.88 8.75
Nilphamari 2.29 3.33 5.83 1.04 6.04 10.63
Panchagarh 2.29 3.13 8.96 0.63 7.08 10.42
Rangpur 2.71 4.17 7.71 1.04 5 9.38
Thakurgaon 1.67 4.38 10.21 1.04 5 11.67
Average 2.08 4.17 8.31 1.07 5.21 10.31
Sylhet Habiganj 2.08 2.29 5.63 3.96 2.29 8.13
Maulvibazar 1.25 3.96 9.17 3.33 3.54 7.71
Sunamganj 3.33 2.08 5.42 3.13 1.88 7.5
Sylhet 0.63 4.79 9.58 2.08 4.38 7.71
(continued)
Table 2 (continued)
Average 1.82 3.28 7.45 3.13 3.02 7.76
Fig. 2 Spatial distribution of the frequency (in %) of short-term dry periods (a) and frequency of
short-term wet periods (b) of different intensities (extreme, severe and moderate)
that year against the December–January–February precipitation totals for all years.
Short-term analysis indicates a greater likelihood of extreme droughts in the bulk
of districts in the country’s northernmost area of Rangpur and east to south-eastern
Chittagong region, according to the 3-month SPI. Severe droughts are prevalent in
northern Bangladesh, while moderate droughts affect the middle to south-western
parts of the country. Short-term severe to extreme wet episodes are more prevalent
134 S. Mallik
Table 3 The frequency of occurrences of short-term dry and wet periods of various intensities in
different districts of Bangladesh between the years 1981 and 2020
Barisal Barguna 4.17 16.67 45 9.17 22.5 35
Barisal 5 20 38.33 8.33 23.33 38.33
Bhola 5 15.83 47.5 7.5 15.83 45
Jhalokati 5 15.83 46.67 9.17 22.5 33.33
Patuakhali 5.83 13.33 49.17 10 20.83 32.5
Pirojpur 3.33 14.17 47.5 11.67 21.67 30.83
Average 4.72 15.97 45.7 9.31 21.11 35.83
Chittagong Bandarban 5 11.67 47.5 7.5 13.33 33.33
Brahamanbaria 6.67 15.83 36.67 11.67 14.17 36.67
Chandpur 5.83 15.83 42.5 10 18.33 36.67
Chittagong 6.67 16.67 35.83 8.33 15.83 37.5
Cumilla 6.67 17.5 29.17 8.33 21.67 34.17
Coxs Bazar 5.83 12.5 39.17 6.67 16.67 28.33
Feni 5.83 20.83 39.17 5.83 19.17 48.33
Khagrachhari 5.83 16.67 33.33 7.5 21.67 38.33
Lakshmipur 5.83 14.17 44.17 6.67 20 36.67
Noakhali 5 17.5 48.33 5 20 41.67
Rangamati 7.5 12.5 37.5 7.5 19.17 33.33
Average 6.06 15.61 39.39 7.73 18.18 36.82
Dhaka Dhaka 4.17 16.67 41.67 11.67 15 44.17
Faridpur 2.5 15.83 43.33 11.67 19.17 34.17
Gazipur 5.83 15.83 40 9.17 20.83 40.83
Gopalganj 2.5 19.17 44.17 13.33 15.83 34.17
Jamalpur 6.67 21.67 38.33 8.33 14.17 38.33
Kishoreganj 6.67 16.67 40 9.17 16.67 38.33
Madaripur 3.33 16.67 42.5 9.17 20.83 40
Manikganj 3.33 18.33 42.5 10 20.83 45
Munshiganj 5.83 15.83 42.5 11.67 16.67 40.83
Mymensingh 5 20 34.17 4.17 16.67 41.67
Narayanganj 5 15.83 45.83 8.33 17.5 44.17
Narsingdi 6.67 15 37.5 10 17.5 36.67
Netrakona 5 18.33 41.67 6.67 11.67 45.83
Rajbari 3.33 15.83 43.33 9.17 19.17 36.67
Shariatpur 4.17 19.17 39.17 10.83 19.17 34.17
Sherpur 6.67 19.17 40.83 6.67 16.67 40.83
(continued)
Table 3 (continued)
Tangail 3.33 20.83 45 5.83 22.5 38.33
Average 4.71 17.7 41.32 9.17 17.7 39.66
Khulna Bagerhat 1.67 15 51.67 13.33 14.17 33.33
Chuadanga 3.33 17.5 40.83 10.83 20.83 33.33
Jessore 3.33 19.17 40.83 10 18.33 40.83
Jhenaidah 3.33 20.83 38.33 10 21.67 31.67
Khulna 1.67 13.33 48.33 10.83 17.5 34.17
Kushtia 5 17.5 45 10 19.17 32.5
Magura 3.33 13.33 44.17 10.83 19.17 39.17
Meherpur 5 18.33 36.67 12.5 17.5 30
Narail 1.67 16.67 50.83 12.5 15 39.17
Satkhira 3.33 11.67 44.17 12.5 15.83 33.33
Average 3.17 16.33 44.08 11.33 17.92 34.75
Rajshahi Bogra 5 22.5 38.33 7.5 13.33 41.67
Joypurhat 7.5 19.17 37.5 6.67 19.17 35.83
Naogaon 6.67 22.5 31.67 6.67 20 38.33
Natore 5 18.33 48.33 7.5 15 43.33
Nawabganj 5.83 19.17 39.17 9.17 17.5 31.67
Pabna 3.33 16.67 46.67 10 21.67 32.5
Rajshahi 4.17 21.67 39.17 7.5 18.33 34.17
Sirajganj 4.17 15.83 45 5.83 22.5 35.83
Average 5.21 19.48 40.73 7.61 18.44 36.67
Rangpur Dinajpur 5.83 22.5 35 9.17 10.83 38.33
Gaibandha 5.83 17.5 40.83 7.5 16.67 41.67
Kurigram 7.5 18.33 35.83 6.67 17.5 28.33
Lalmonirhat 8.33 18.33 33.33 10.83 9.17 45.83
Nilphamari 7.5 20.83 31.67 10.83 14.17 37.5
Panchagarh 4.17 19.17 36.67 10.83 11.67 39.17
Rangpur 8.33 18.33 35.83 6.67 15.83 34.17
Thakurgaon 6.67 17.5 33.33 7.5 18.33 35.83
Average 5.02 17.31 41.08 9.02 17.87 37.22
Sylhet Habiganj 5 19.17 37.5 9.17 17.5 35.83
Maulvibazar 4.17 18.33 35 10 16.67 34.17
Sunamganj 5 16.67 45.83 6.67 12.5 46.67
Sylhet 3.33 19.17 48.33 9.17 14.17 37.5
(continued)
136 S. Mallik
Table 3 (continued)
Average 4.38 18.34 41.67 8.75 15.21 38.54
in Bangladesh’s central and south-western areas, compared to moderate wet periods

which are prevalent in the central and north-eastern regions.
Additionally, there is a notable change in the frequency percentages—they
increase for both dry and wet periods of all intensities as the timescale decreases.
Tables 2 and 3 illustrate this by comparing the frequency percentages for various
events in the 3- and 12-month SPI. By identifying critical districts i.e., the districts
highlighted in red in the generated maps, short-term solutions to meet agricul-
tural objectives, such as achieving optimal soil moisture and minimizing agricul-
tural drought risks, may be proposed for those regions. It is prudent to take appro-
priate steps early in response to moderate dry/wet episodes that have the potential to
deteriorate into severe occurrences over longer time frames.
3.3 Spatial Distribution of the Total Number of Dry/Wet

Events in Bangladesh Since 1981
The districts with the largest number of dry/wet events, regardless of intensity, are
depicted in Fig. 3. It is apparent that nearly the whole western Bangladesh is prone to
long-term dry episodes, but short-term dry episodes are more frequent in the southern
districts of the Barisal division and portions of the north and north-eastern districts of
Sylhet and Rajshahi division. Similarly, whereas the northern and southern sections
of the nation are more susceptible to long-term wet periods, the central region of
Fig. 3 Spatial distribution of the total number of long-term (a) and short-term (b) dry months and
wet months in Bangladesh between 1981 and 2020
Bangladesh is more prone to short-term wet periods. It is worth noting that while
certain districts (e.g., Dhaka) are prone to frequent short-term wet periods, they are
one of the least affected areas in terms of long-term wet periods. This confirms the
spatial variability of dry/wet periods in terms of timescales.
3.4 Evaluation of SPI-3 and SPI-12 in 40 years
SPI time series was constructed at different time intervals of three and twelve months
to assess their potential utility for identifying dry/wet periods and monitoring the
risk of extreme events. Figure 5 shows the SPI time series for the eastern Sylhet and
northern Rangpur regions from 1981 to 2020 on 3- and 12-month time periods, as
well as the occurrence of past major flood and drought events in the region.
Dry/wet seasons vary according to timescale. When timescale is small, the SPI
oscillates frequently above and below zero. On longer periods, the SPI varies slowly
as precipitation changes. When a 12-month period is used, the drought episodes
are more concentrated and persist longer than the more dispersed 3-month drought
events—they occur more frequently and last less time. This is a feature that has
been seen in a variety of other locations: McKee et al. (1993) in Colorado, USA,
Lloyd-Hughes and Saunders (2002) in Yorkshire, UK, and Szalai and Szinell (2000)
in Szarvas, Hungary.
Figure 4 depicts the variability of the SPI in the Sylhet and Rangpur districts over
3- and 12-month intervals from 1981 to 2020. The findings confirm the assertions
made previously. SPI demonstrates a high frequency of shift between dry and rainy
periods over short time frames. Dry and wet phases shift less frequently and last
longer as timescales increase.
Longer timescales (SPI of 12 months or more) are better for detecting historically
major drought and flood incidents. Drought episodes lasting 12 months or longer
(SPI-12 < −1.5) were recorded in Sylhet in 1994, 2007, and 2019 and in Rangpur in
1994, 2007, and 2014. In Sylhet, severe to extreme flood episodes occurred in 1988,
2004, 2010, and 2017; while in Rangpur, severe to extreme flood events occurred
in 1984, 1988, 1998, and 2020. Rangpur had a succession of extremely dry years
from 2007 to 2015, with the detrimental impacts of droughts reaching an apex in
2014. During the study period, Sylhet, on the other hand, had interchanging years of
dry/wet periods.
On shorter timescales, the frequent seasonal and inter-annual precipitation fluc-
tuations are more apparent, resulting in a large number of events and, thus, a greater
frequency. On a short timescale, the class of moderate events has the highest monthly
frequency, at 40–45% for moderate dry episodes and 35–40% for moderate wet
episodes. On the contrary, 8–12% for moderate dry and 6–10% for moderate wet
episode frequency are observed in terms of longer timescale, both of which are
highest among the classes. The shorter timescale SPI values are more scattered and
fluctuate rapidly, whereas the longer timescale SPI values are localized in a series of
138 S. Mallik
Fig. 4 The temporal evaluation of SPI-3 and SPI-12 averaged over districts of (a) Sylhet and
(b) Rangpur
consecutive months within a year or the next few years (e.g., the Rangpur district’s
dry years of 2007–2015).
By translating the observed number of dry/wet events to the number of occurrences
in 100 years using Eq. (2), one can anticipate for the district of Sylhet approximately
15 droughts and 14 flood events of varying intensities lasting 12 months (Table 2),
as well as approximately 71 droughts and 61 flood episodes of varying intensities
lasting 3 months (Table 3). Using a similar method, comparable estimations can be
generated for other districts as well. The aforementioned data, derived from the SPI
study, provides extremely useful information about the local climates found across
Bangladesh. For many years, it has been asserted that this is a region prone to droughts
and floods, but only the SPI for various timescales and intensities allows for the
visualization of the spatial link between the number of months with droughts/floods
for various timescales, frequency, and length.
4 Conclusions
Since point-based observations limit spatiotemporal analysis of precipitation, effec-

tive drought monitoring is hindered. High-resolution satellite precipitation data prod-
ucts, when combined with in situ observations, have been shown to be an excel-
lent alternative data source for spatiotemporal drought analysis. The open-source
CHIRPS data products are accessible at multiple temporal scales and have a great
potential for drought and flood detection and evaluation in near real-time. This study
evaluated the district-wide dry/wet event characteristics for Bangladesh from 1981
to 2020 using the CHIRPS data generated SPI. The findings of this study show that
the standardized precipitation index SPI is extremely useful in identifying and char-
acterizing local dry and wet periods. Since it just requires precipitation data, the SPI
is straightforward to use. It is feasible to determine the frequency of droughts and
floods that are the result of rainfall anomalies in previous periods using different
timescales for which it is computed. The study looked at the number of months in
which drought/flood events were recognized at the 3- and 12-month timescales and
found that, depending on the timing and intensity of the event, there is a significant
geographical diversity of dry/wet period recurrence frequency across Bangladesh.
Based on the three classes of intensities, the highest frequencies of long-term dry
periods were observed in divisions of Rangpur (Extreme-2.08%), Rajshahi (Severe-
6.96%), and Khulna (Moderate-12.02%), whereas long-term wet periods were
prevalent in divisions of Dhaka (Extreme-3.32%) and Rangpur (Severe-5.21% and
Moderate-10.31%). Moreover, short-term analysis depicted a different scenario, with
the Chittagong division (Extreme-6.06%), the Rajshahi division (Severe-19.48%),
and the Barisal division (Moderate-45.7%) observed as the most frequently affected
by short-term dry periods; and the Khulna division (Extreme-11.33%), the Barisal
division (Severe-21.11%), and the Dhaka division (Moderate-39.66%) observed as
the most short-term flood-prone divisions. Using previous rainfall data, the current
study aids in comprehending severe dry/wet occurrences in different locations of
the country. Furthermore, the findings may aid in the planning of essential steps to
control regional drought and flood, reducing the negative effects of extreme weather
occurrences across Bangladesh.
References
Adams, James. 2017. Climate_indices, an Open Source Python Library Providing Reference
Implementations of Commonly Used Climate Indices. https://github.com/monocongo/climate_i
ndices.
Ahmed, Kamal, Shamsuddin Shahid, Eun-Sung. Chung, Xiao-jun Wang, and Sobri Bin Harun.
2019. Climate Change Uncertainties in Seasonal Drought Severity-Area-Frequency Curves: Case
of Arid Region of Pakistan. Journal of Hydrology 570: 473–485.
Alamgir, Mahiuddin, Shamsuddin Shahid, Manzul Kumar Hazarika, Syams Nashrrullah, Sobri
Bin Harun, and Supiah Shamsudin. 2015. Analysis of Meteorological Drought Pattern During
Different Climatic and Cropping Seasons in Bangladesh. JAWRA Journal of the American Water
Resources Association 51 (3): 794–806.
Arnell, Nigel W. 1999. Climate Change and Global Water Resources. Global Environmental Change
9: S31-49.
Bangladesh Bureau of Statistics (BBS). 2015. Disaster-Related Statistics 2015 Climate Change and
Natural Disaster Perspectives. Ministry of Planning, Government of the People’s Republic of
Bangladesh: Dhaka, Bangladesh.
140 S. Mallik
Bangladesh Bureau of Statistics (BBS). 2021. Yearbook of Agricultural Statistics-2020. Ministry

of Planning, Government of the People’s Republic of Bangladesh: Dhaka, Bangladesh.
Basher, Md Abul, Mathew Alexander Stiller-Reeve, A.K.M. Saiful Islam, and Scott Bremer. 2018.
Assessing Climatic Trends of Extreme Rainfall Indices over Northeast Bangladesh. Theoretical
and Applied Climatology 134 (1): 441–452.
Bechle, Matthew J., Dylan B. Millet, and Julian D. Marshall. 2013. Remote Sensing of Expo-
sure to NO2: Satellite Versus Ground-Based Measurement in a Large Urban Area. Atmospheric
Environment 69: 345–353.
Cordell, Dana, Jan-Olof. Drangert, and Stuart White. 2009. The Story of Phosphorus: Global Food
Security and Food for Thought. Global Environmental Change 19 (2): 292–305.
Du, Juan, Jian Fang, Wei Xu, and Peijun Shi. 2013. Analysis of Dry/Wet Conditions Using the
Standardized Precipitation Index and Its Potential Usefulness for Drought/Flood Monitoring
in Hunan Province, China. Stochastic Environmental Research and Risk Assessment 27 (2):
377–387.
Edwards, Daniel C. 1997. Characteristics of 20th Century Drought in the United States at Multiple
Time Scales. Air Force Inst of Tech Wright-Patterson AFB OH.
Funk, Chris, Pete Peterson, Martin Landsfeld, Diego Pedreros, James Verdin, James D. Rowland,
Bo E. Romero, Gregory Husak, Joel Michaelsen, and Andrew Verdin. 2014. A Quasi-Global
Precipitation Time Series for Drought Monitoring. US Geological Survey Data Series 832 (4):
1–12.
Funk, Chris, Pete Peterson, Martin Landsfeld, Diego Pedreros, James Verdin, Shraddhanand Shukla,
Gregory Husak, James Rowland, Laura Harrison, and Andrew Hoell. 2015. The Climate Hazards
Infrared Precipitation with Stations—A New Environmental Record for Monitoring Extremes.
Scientific Data 2 (1): 1–21.
Gao, Feng, Yuhu Zhang, Xiulin Ren, Yunjun Yao, Zengchao Hao, and Wanyuan Cai. 2018. Evalu-
ation of CHIRPS and Its Application for Drought Monitoring over the Haihe River Basin, China.
Natural Hazards 92 (1): 155–172.
“Glossary of Terms.” 2014. Scandinavian Journal of Public Health 42 (14_suppl): 178–190. https://
doi.org/10.1177/1403494813515131.
Green, Colin. 2010. Towards Sustainable Flood Risk Management. International Journal of Disaster
Risk Science 1 (1): 33–43.
IEA. 2009. World Energy Outlook 2009 – Analysis - IEA. https://www.iea.org/reports/world-ene
rgy-outlook-2009.
Kafy, Abdulla - Al, Abdullah Al Rakib, Kaniz Shaleha Akter, Zullyadini A. Rahaman, Abdullah-Al
Faisal, Saumik Mallik, N.M. Refat Nasher, Md. Iquebal Hossain, and Md. Yeamin Ali. 2021.
Monitoring the Effects of Vegetation Cover Losses on Land Surface Temperature Dynamics
Using Geospatial Approach in Rajshahi City, Bangladesh. Environmental Challenges: 100187.
https://doi.org/10.1016/j.envc.2021.100187.
Keyantash, John, and John A. Dracup. 2002. The Quantification of Drought: An Evaluation of
Drought Indices. Bulletin of the American Meteorological Society 83 (8): 1167–1180.
Khan, Sadiq I., Yang Hong, Jonathan J. Gourley, Muhammad U. Khattak, and Tom De Groeve.
2014. Multi-Sensor Imaging and Space-Ground Cross-Validation for 2010 Flood along Indus
River, Pakistan. Remote Sensing. https://doi.org/10.3390/rs6032393.
Kim, Do-Woo, Hi-Ryong Byun, and Ki-Seon Choi. 2009. Evaluation, Modification, and Application
of the Effective Drought Index to 200-Year Drought Climatology of Seoul, Korea. Journal of
Hydrology 378 (1–2): 1–12.
Krysanova, Valentina, Tobias Vetter, and Fred Hattermann. 2008. Detection of Change in Drought
Frequency in the Elbe Basin: Comparison of Three Methods. Hydrological Sciences Journal 53
(3): 519–537.
Livada, I., and V.D. Assimakopoulos. 2007. Spatial and Temporal Analysis of Drought in Greece
Using the Standardized Precipitation Index (SPI). Theoretical and Applied Climatology 89 (3):
143–153.
McKee, Thomas B., Nolan J. Doesken, and John Kleist. 1993. The Relationship of Drought
Frequency and Duration to Time Scales. In Proceedings of the 8th Conference on Applied
Climatology 17: 179–183. Boston.
Min, Seung-Ki, Won-Tae Kwon, E-Hyung Park, and Youngeun Choi. 2003. Spatial and Temporal
Comparisons of Droughts over Korea with East Asia. International Journal of Climatology: A
Journal of the Royal Meteorological Society 23 (2): 223–233.
Mohsenipour, Morteza, Shamsuddin Shahid, Eun-sung Chung, and Xiao-jun Wang. 2018. Changing
Pattern of Droughts during Cropping Seasons of Bangladesh. Water Resources Management 32
(5): 1555–1568.
Molden, David. 2013. Water for Food Water for Life: A Comprehensive Assessment of Water
Management in Agriculture. Routledge.
Mondol, Md, Anarul Haque, Iffat Ara, and Subash Chandra Das. 2017. Meteorological Drought
Index Mapping in Bangladesh Using Standardized Precipitation Index during 1981–2010.
Advances in Meteorology 2017.
Moradi, H.R., M. Rajabi, and M. Faragzadeh. 2011. Investigation of Meteorological Drought
Characteristics in Fars Province, Iran. CATENA 84 (1–2): 35–46.
Nafarzadegan, A.R., M. Rezaeian Zadeh, M. Kherad, H. Ahani, A. Gharehkhani, M.A. Karam-
poor, and M.R. Kousari. 2012. Drought Area Monitoring During the Past Three Decades in Fars
Province, Iran. Quaternary International 250: 27–36.
Pandey, Varsha, Prashant K. Srivastava, R.K. Mall, Francisco Munoz-Arriola, and Dawei Han. 2020.
Multi-Satellite Precipitation Products for Meteorological Drought Assessment and Forecasting
in Central India. Geocarto International, 1–20.
Parry, Martin, Osvaldo Canziani, Jean Palutikof, Paul van der Linden, and C.E. Hanson. 2007.
Intergovernmental Panel on Climate Change Contribution of Working Group I to the Fourth
Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge and New
York: Cambridge University Press.
Piccarreta, Marco, Domenico Capolongo, and Federico Boenzi. 2004. Trend Analysis of Precipita-
tion and Drought in Basilicata from 1923 to 2000 within a Southern Italy Context. International
Journal of Climatology: A Journal of the Royal Meteorological Society 24 (7): 907–922.
Quiring, Steven M., and Timothy N. Papakryiakou. 2003. An Evaluation of Agricultural Drought
Indices for the Canadian Prairies. Agricultural and Forest Meteorology 118 (1–2): 49–62.
Rimi, Ruksana H., Karsten Haustein, Myles R. Allen, and Emily J. Barbour. 2019. Risks of Pre-
Monsoon Extreme Rainfall Events of Bangladesh: Is Anthropogenic Climate Change Playing a
Role. Bulletin of the American Meteorological Society 100: S61–65.
Rosenzweig, Cynthia, Ana Iglesius, Xiao-Bing Yang, Paul R. Epstein, and Eric Chivian. 2001.
Climate Change and Extreme Weather Events-Implications for Food Production, Plant Diseases,
and Pests.
Shahid, Shamsuddin, and Houshang Behrawan. 2008. Drought Risk Assessment in the Western Part
of Bangladesh. Natural Hazards 46 (3): 391–413. https://doi.org/10.1007/s11069-007-9191-5.
Shrestha, Narayan K., Faisal M. Qamer, Diego Pedreros, M.S.R. Murthy, Shahriar Md Wahid, and
Mandira Shrestha. 2017. Evaluating the Accuracy of Climate Hazard Group (CHG) Satellite
Rainfall Estimates for Precipitation Based Drought Monitoring in Koshi Basin, Nepal. Journal
of Hydrology: Regional Studies 13: 138–151.
Stagge, James H., Lena M. Tallaksen, Lukas Gudmundsson, Anne F. Van Loon, and Kerstin Stahl.
2015. Candidate Distributions for Climatological Drought Indices (SPI and SPEI). International
Journal of Climatology 35 (13): 4027–4040.
Tao, Hui, Thomas Fischer, Yan Zeng, and Klaus Fraedrich. 2016. Evaluation of TRMM 3B43
Precipitation Data for Drought Monitoring in Jiangsu Province, China. Water 8 (6): 221.
Thom, Herbert C S. 1958. A Note on the Gamma Distribution. Monthly Weather Review 86 (4):
117–122.
Vergni, L., and F. Todisco. 2011. Spatio-Temporal Variability of Precipitation, Temperature and
Agricultural Drought Indices in Central Italy. Agricultural and Forest Meteorology 151 (3):
301–313.
142 S. Mallik
WMO. 2012. Standardized Precipitation Index User Guide. https://library.wmo.int/doc_num.php?

explnum_id=7768.
Wu, Hong, Mark D. Svoboda, Michael J. Hayes, Donald A. Wilhite, and Fujiang Wen. 2007.
Appropriate Application of the Standardized Precipitation Index in Arid Locations and Dry
Seasons. International Journal of Climatology: A Journal of the Royal Meteorological Society
27 (1): 65–79.
Wu, Wenqi, Yungang Li, Xian Luo, Yueyuan Zhang, Xuan Ji, and Xue Li. 2019. Performance
Evaluation of the CHIRPS Precipitation Dataset and Its Utility in Drought Monitoring over
Yunnan Province, China. Geomatics, Natural Hazards and Risk 10 (1): 2145–2162.
Yuan, Wen-Ping., and Guang-Sheng. Zhou. 2004. Comparison Between Standardized Precipitation
Index and Z-Index in China. Chinese Journal of Plant Ecology 28 (4): 523.
Zhong, Ruida, Xiaohong Chen, Chengguang Lai, Zhaoli Wang, Yanqing Lian, Haijun Yu, and
Xiaoqing Wu. 2019. Drought Monitoring Utility of Satellite-Based Precipitation Products Across
Mainland China. Journal of Hydrology 568: 343–359.
Indigenous Knowledge and Practices
of the Small Ethnic Communities
of Asia–Pacific Island Countries
in Facing Hydro-Meteorological Hazards
Mahfuzul Haque
Abstract Tsunami of 2004 brought with it unprecedented damages to the lives and
livelihoods of people in 12 Indian Ocean nations. This was by far, the worst disaster
in the recent past, resulting in a high death toll and a huge devastation. Studies
conducted during the post-tsunami period among the coastal communities of the
island nations in South Asia and the Pacific found survival strategies of the local
community in the face of disasters. This paper examined resilience of the coastal and
small island communities in Indonesia, the Philippines, Timor-Leste and Solomon
Islands in the Asia–Pacific; Andaman and Nicobar Islands in the Indian Ocean;
and Bangladesh coast in the Bay of Bengal in the face of earthquake, tsunami and
hazards like cyclones, floods and tidal surges. Based on animal behavior, activities of
celestial bodies, environment; material culture; traditional, religious and faith-based
beliefs and practices, the local community was able to develop a coping mechanism
for themselves with these hydro-meteorological disasters. The paper argues that as
the indigenous communities practice their traditional knowledge, they could save
themselves from the earthquake and tsunami-related disasters. The paper further
argues that they have been living with natural disasters over generations based on
their indigenous knowledge and practices.
Keywords Andaman and Nicobar Island · Bangladesh coast · Ethnic

communities · Indigenous knowledge and practices · Tsunami
1 Introduction
According to UNESCO’s program on Local and Indigenous Knowledge Systems

(LINKS), local and indigenous knowledge refers to the understandings, skills and
philosophies developed by societies with long histories of interaction with their
natural surroundings (Hiwasaki et al. 2014). This definition states that local and
indigenous knowledge is one that is dynamic. The term local and indigenous
M. Haque (B)
Adjunct Faculty, Department of Sociology, Bangladesh University of Professionals, Dhaka,
Bangladesh
144 M. Haque
knowledge is comparable to local and traditional knowledge, indigenous technical

knowledge and endogenous knowledge.
Functional knowledge of local people is generally known as Indigenous Knowl-
edge (IK) living in a particular ethno-cultural and agro-ecological condition. Indige-
nous knowledge is developed through experience sharing and normally passes from
one generation to another orally, and it remains in operation in all aspects of commu-
nity life and continues to be there in unwritten forms (Amin 2000; Haque 2013). It is
often mentioned that unwritten knowledge is vulnerable to being lost; local wisdom
has a tendency to remain present in the villages, old towns, markets and other places.
IK generally sticks around in a holistic way in their knowledge, beliefs and percep-
tions, mainly among the people living in rural environments (Warner 1991; Sillitoe
et al. 1998). According to Walker et al. (1991), this is a knowledge held collectively
by a specific community.
Generally, indigenous knowledge is not codified or written in formal language.
IK mostly remains in verbal form and passes on from one generation to another. It is
increasingly felt that IK is a resource that should be promoted and preserved in order
to complement scientific knowledge while developing appropriate plans and policies
for people in the rural areas. IK refers to knowledge, values, concepts, perceptions
and beliefs of a specific community. It is essentially a diverse and multi-faceted
knowledge system with a distinct language. Sillitoe (2000) stated that contrary to
“top down” interventions, IK has followed a “bottom-up” or “grassroots” process.
Generally found in rural areas in unwritten form and in informal discussions, IK could
be distinguished by scientific knowledge available in written and codified language
and literature.
The term “indigenous” is synonymous with “traditional” and “local”, distin-
guishing this knowledge from that of scientific knowledge developed by formal
research studies undertaken by education institutions (Mustafa 2000; Haque 2019).
IK continues to be used by rural people in natural resource management, conserva-
tion of fisheries, livestock and healthcare practices. Sillitoe et al. (1998) stated that
indigenous knowledge relates to any knowledge held collectively by a population
mostly inhabiting in rural areas. IK may incorporate knowledge of any kind including
that of socio-cultural practices and natural resource management.
Importance of local knowledge and practices cannot be avoided. It is to be noted
that knowledge and practices of the local community cannot be called as primitive
or unscientific. The local community, however, is exposed to modern knowledge.
They have not disowned their indigenous knowledge and practices assimilated over
generations. It is true that local knowledge is eroding fast and much has been lost with
the modernization and rapid spread of foreign technology introduced from outside.
The development practitioners are of the opinion that indigenous knowledge needs
to be promoted and documented before it goes into oblivion.
McAdoo et al. (2008) stated that there are a number of international agreements
that recognize the value of indigenous knowledge for sustainable development.
This includes the Rio Declaration 1992; Agenda 21, 1992; Convention on Biolog-
ical Diversity, 1992; World Summit on Sustainable Development, Johannesburg,
Indigenous Knowledge and Practices of the Small Ethnic … 145
2002; and the work of International Decade (1995–2004) for the World’s Indigenous
Peoples.
In line with the Hyogo Framework for Action 2005–2015 and Sendai Framework
for Disaster Risk Reduction 2015–2030, national disaster risk reduction strategies
are to incorporate the local knowledge of the indigenous communities practiced
over generations. One of the five main drivers of the Hyogo Framework was strong
community engagement for effective disaster risk reduction. Among the priority
actions, the Framework observed that people are to be made aware of indigenous
knowledge and traditional practices for risk reduction and mitigation. On the other
hand, the Sendai Framework, 2015–2030 emphasized on “indigenous knowledge and
practices and coping strategy of the local community in facing challenges of natural
disaster”. Emphasizing on the role of the stakeholders, it said that the indigenous
peoples, through their experience and traditional knowledge, provide an important
contribution to the plans and policies concerning disaster risk reduction.
The hydro-meteorological hazard discussed in this article is atmospheric, hydro-
logical or oceanographic in nature, that may cause loss of life, injury, prop-
erty damage, loss of livelihoods and services, social and economic disruption
and environmental damage. The hazards discussed in this article are caused by
hydro-meteorological events like floods, cyclones, tidal surges and tsunami.
With a brief description on indigenous knowledge and practices, the paper delib-
erates on survival strategies of these coastal and small ethnic communities. The
paper examined how their age-old beliefs were found handy in time of disasters, like
earthquake and tsunami. The paper suggests that indigenous knowledge and prac-
tices need to be integrated with science before it can be used in policies, education
and actions related to hydro-meteorological hazards. This article is divided into five
sections. The introduction deliberates on the definitional issues of indigenous knowl-
edge and practices and the emphasis given by the Hyogo and Sendai Frameworks on
incorporating indigenous knowledge and practices in disaster management policies
and plans. The introduction is followed by a series of case studies concerning the
experiences of the coastal and small ethnic communities of the Asia–Pacific Islands;
Andaman and Nicobar Islands and the Bangladesh Coasts. The last section concludes
that indigenous knowledge and practices need to be integrated with science before it
can be used in policies, education and actions related to disaster risk reduction and
climate change.
2 Methodology
The paper is based on secondary sources of literature, books and journals avail-
able in print and electronic media. Interviews were also held with some authors on
indigenous knowledge and practices, both at home and abroad, in order to under-
stand the link between disaster and its understanding by the coastal and small ethnic
communities, based on their age-old beliefs. Efforts were taken to obtain confir-
mation regarding this delicate linkage through communication with some relevant
146 M. Haque
authors. Views and comments received from the audience present in the conference
were further checked and incorporated prior to finalizing the paper.
3 Asia–Pacific Islands
Primary hydro-meteorological hazards faced by the communities in Asia–Pacific

island nations are typhoons, storms, tidal surges, heavy rainfall, floods and landslides.
Coastal communities have also observed climatic changes, such as warmer days and
nights, changes in rainfall patterns and more frequent and stronger cyclonic storms.
These are mainly seen in the Philippines. In Indonesia, more frequent monsoons,
tropical cyclones, coastal erosion and land subsistence are noticeable. The primary
problems in Timor-Leste are tropical cyclones with heavy rainfall, and prolonged dry
and extended rainy seasons resulting in floods, landslides and droughts (Hiwasaki
et al. 2014). In addition to earthquakes and tsunamis, there have been incidences of
volcanic eruptions from some of the active volcanoes that are present here.
Coastal and small island communities, over the generations, have been observing
changes in the environment and have gathered a wealth of knowledge and practices
closely related to hydro-meteorological hazards. The local people are able to closely
observe and monitor changes in their environment by looking at the seas, clouds,
animals, plants and insects. The changes of position in celestial bodies, like the moon,
sun, and stars also help them understand the advent of hydro-meteorological hazards.
These small island communities also suffer from food crises as they get isolated from
the mainland, and communication is disrupted during hazardous situations.
Hiwasaki et al. (2014) stated that the communities predict heavy rainfall or strong
winds by observing clouds, waves, winds, sun and the stars. Their local and indige-
nous knowledge is based on the (a) observations of animal behavior; (b) observations
of celestial bodies; (c) observations of the environment; (d) material culture; and (e)
traditional and faith-based beliefs and practices. They look for changes in the texture
of clouds (thin or thick), color (white, dark, yellow or red), location (over the moun-
tains or the sea) and movement (to and from the coast), including speed (fast) and
direction (vertical or horizontal). They observe how the sky changes color, as well
as the direction and height of the waves. The direction of winds and temperature; the
position and size of the sun; and visibility and constellations of stars are signs that
communities commonly look out for. The islanders shared that they can smell a foul
odor from the sea signaling that a storm or typhoon is coming.
In some Pacific coastal islands, people watch movements of animals, insects and
plants for predicting hazards. In the islands of Timor-Leste, leaches and caterpillars
are seen before typhoons. When banana tree leaves fall to the ground without strong
winds, people in the Philippines prepare for storms or typhoons (Hiwasaki et al.
2014). In Timor-Leste and Indonesian islands, birds, usually migratory, are important
signs of changing seasons and duration. In the islands of Philippines, various animals
are used to forecast hazards as the rays jump consecutively in the sea in summer,
the fast movement of sea snakes and hermit crabs going inland or climbing up trees.
These observations are also considered as signs of other hazards like flooding and
landslides. Similar to predictions and warnings, islanders have also developed ways
to prevent or mitigate such hazards, using local materials. They plant local trees to
safeguard their houses from the storm. They have developed various ways to ensure
food security during storms and droughts. They use salt to preserve fish and often
smoke-dry them. People used to consume locally available fruits, tuber and roots in
the aftermath of a disaster when external relief took time to come. In order to protect
their houses from storms and strong winds, they use local structures, materials and
plants.
In Solomon Islands in 2007, an earthquake of magnitude 8.1 hit the islands’
western province. The earthquake caused severe damages to structures in the region.
Coral reefs were damaged, and the earthquake triggered landslides, caused uplift
and subsidence and generated a sizable tsunami resulting in the death of 52 people
(McAdoo et al. 2008). There was less than a three-minute separation time between
the earthquake and the tsunami in the Solomon Islands, and the indigenous peoples
based on their traditional knowledge and practices were still able to survive. The
number of casualties would have been higher had the local community not applied
their indigenous knowledge.
What happened was that the water of the coral lagoon was drained out when the
shaking stopped exposing the seafloor. The tsunami advanced within ten minutes
later. The following two or three waves hit the area like high tides. Expecting the
advance of the incoming tsunami, the indigenous people of the island moved toward
the steep hills, and the village elders and heads of household ensured safety of the
children and made proper evacuation anticipating the approach of another big wave.
The people, at least 50, who died during the Solomon Islands tsunami, were mostly
immigrants from the mainland with no such memory. Such localized knowledge
related to natural disasters could also be used effectively during early warning and
response.
4 Andaman and Nicobar Islands
The Andaman and Nicobar Islands in the midst of the Indian Ocean, some 1,200 km
off the eastern coast of India, were severely devastated by an earthquake of 9.0
magnitude followed by tsunami waves during the early hours of 26 December 2004.
Out of 572 islands, only 38 were able to live (Sekhsaria 2017). These islands are
inhabited by some primitive tribal groups, who came from Africa some 20,000 years
ago. With various “development” activities like urbanization, tourism, moderniza-
tion, these primitive tribes gradually lost their culture and identity. Now there are
only 92 Ongees, 43 Great Andamanese, 350 Jarawa, 100 Sentinelese and over 250
Shompens (Reddy 2018), who constitute the “Particularly Vulnerable Tribal Groups”
of Andaman and Nicobar Islands.
During the 26 December 2004 tsunami, the southern islands of the archipelago,
the Nicobar Islands were worst hit (Reddy 2013; Sekhsaria 2014). Geologically,
148 M. Haque
the Andaman and Nicobar Islands represent the highest peaks of an under-water
mountain range. Of the 3,513 people reported dead and missing, only 64 were from
the Andaman group of islands, the remaining 3,449 were from various southern
islands in Nicobar (Reddy 2018). It is to be noted that these islands were about
180 km from the epicenter of the earthquake in Aceh province in Sumatra (Singh
et al. 2018; Sekhsaria 2017). The Sumatra–Andaman earthquake, which caused the
2004 Indian Ocean tsunami, is estimated to have released energy equivalent to 23,000
Hiroshima-type atomic bombs. In Banda Aceh, the landmass closest to the quake’s
epicenter, tsunami waves topped 100 feet. The earthquake-induced tsunami affected
the areas in Malaysia, Myanmar and Thailand. The giant wave later reached the
coasts of Bangladesh, India, Maldives and Sri Lanka. Hours later, it reached the
shores of the Seychelles, Kenya, Somalia and Tanzania in East Africa. Although
the final figure varies as per different sources, the total casualty may vary between
250,000 and 300,000 (Shaw 2006).
The deep-ocean earthquake shook the islands for several minutes followed by
eight consecutive tsunami waves. As the islands were mostly flat, the people were
made easy victims of the incoming waves. The tsunami swept across some islands
causing land subsidence with little signs of human habitation. Uprooted trees were
found floating around in the sea, and the coast was clogged with damaged corals.
Various parts of the islands were deformed, and the island was overall reduced in
size. Although, Nicobar Islands sustained a heavy loss of lives and property, the
other primitive tribe, the Ongees in Little Andaman Islands, could however escape
to safety. The island experienced 6–10 m high waves followed by the earthquake.
Soon after the earthquake, about 83 Ongees packed up and ran through the forest to
a safer place in the high hills (Pandya 2005). Their primitive tribal wisdom helped
them to sense the imminent danger of tsunami waves.
They consider earth tremors as frequent events and have been living in such
situations for years. According to them, with the advent of giyangejebey (tsunami),
the water dries up and goes away from the land very quickly, and similar to the motion
of breathing-in and -out of the body, the sea water had to come back very rapidly
with more force. Giyangejebey in Ongee language is a verb meaning “solid earth
becoming fluid like the sea water” (Pandya 2005). It is not only solids melting in to
liquids, but also liquids becoming solids again. They saw the water and knew that
more land would soon be covered with sea and angry spirits would descend down
to hunt them away. They believed that the Ibedangey (ancestral spirits) would come
down to help them if they remained united (Chandi and Andrews 2010).
According to Ongees, the Lololokobey (earthquakes) are frequent and natural.
They associate earthquakes with the creation of dead ancestors crossing over to
another-world and becoming spirits (Pandya 2005). The image of rising water levels,
as in the case of tsunami, has often been referred to as floods, a disaster associated
since mythical times with the destruction of fire by rising water levels. There is a
cultural practice among the primitive tribes in Andaman and Nicobar Islands to save
fire from being extinguished by water, meaning fire should remain burning for their
safety.
Based on their age-old cultural wisdom, the Ongees got concerned on the morning
of the giant tsunami when they saw the rapidly receding water line, something that
signaled to them that the “ancestral spirits” were angry and were shaking the pillar-
like tree trunk on which the sea rests. According to the Ongees tradition, when the
spirits get angry, the wind directions change, and spirits move from sea to forest
during September–October and move from the forest to sea during July to August
(Pandya 2005). In the state of anger, the spirits often hurl down huge boulders taken
from stars at the sea. Earthquakes are experienced by the impact of the boulders
touching the sea surface.
On the morning of tsunami, when the water line receded, the Ongees immediately
gathered on the shores with their baskets, bows and arrows and hurled stones at the
sea, for they wanted the spirits to believe that the community was still left behind
submerged under the rising water level, so that the spirits remained at the spot looking
for those who had hurled stones at them while other spirits continued to shake the
earth. Unlike the settlers, the Ongees did not wait to see aftermaths of the earthquake,
which to them was a regular event. Many lives of the settler communities were lost
because they were looking at the receding sea, oblivious to the fact that the killer
waves were coming from behind.
The Ongees’ idea of living in the future is based on their cultural notion of when
to move and where to move. In fact, the Ongees, as stated by Pandya (2005), usually
move to the interior forest or coastal area and set up traditional residential structures
known as korale (single family shelters) or beraley (communal huts). This practice of
translocation has always made the Ongees adept at packing up essentials and moving
on in response to seasonal changes.
The Sentinelese and Jarawas are the communities, who also continue to use their
knowledge networks for livelihood security. Like others, they are the hunter-gatherer
tribes. The Jarawas know the forests they live in. They are familiar with the medicinal
plants and the use of special herbs for driving away bees from bee-hive (Chandi 2010;
Sekhsaria 2014). It is also learnt that sensing imminent danger of tsunami, the most
isolated and fierce primitive tribe in the North Sentinel island, the Sentinelese were
able to leave their homestead and escape to the highland forests for safety, although
it is not known if they lost property and lives during the incident. Indian helicopters
with relief materials had to abort their mission as they faced a hail of arrows from
these tribes.
5 Bangladesh Coastal Islands and the Haors
The coastal islands of Bangladesh on the Bay of Bengal are periodically vulner-
able to hydro-meteorological disasters like floods, cyclones, tidal surges and salinity
intrusion. Threat of sea level rise due to climate change is also looming. The 710 km
coastal line covering many small and newly emerged charlands (shoals) are subjected
to tropical cyclones and storm surges that regularly devastate vast areas with salinity
intrusion and prolonged inundation, incurring loss of lives and damages to crops. The
150 M. Haque
coastal area comprises of 47,201 sq.km (32% of the country) of Bangladesh within
19 districts in 147 Upazilas (Rasheed 2008). The area is populated by 35 million
people (2005) representing 29% of the total population.
Because of the very proximity with the Bay of Bengal, the coastal population,
especially the charland people have developed through a process of adaptation, many
coping strategies and techniques in line with the local environment. These coping
strategies are based on their local knowledge and practices gathered over the gener-
ations. Unfortunately, much of the knowledge is lost as they remain undocumented.
Nevertheless, people in disaster-prone areas still nurture such knowledge in their
myths, beliefs and traditions (Haque 2019).
The coastal landmass of Bangladesh is periodically subjected to hydro-
meteorological disasters, like cyclones and tidal surges causing loss of lives and prop-
erties. In the last 150 years, 35 devastating cyclones have hit the Bangladesh Coast
(Haque 2019). Generally, April–May and November are the months of cyclones and
associated tidal surges in Bangladesh. Around 330,000 people in the coastal islands of
Bangladesh died as a result of the cyclone and tidal surges of 12 November 1970, and
another 138,882 people died on 29 April 1991 (Haque 2019). Bangladesh recently
faced a number of devastating cyclonic storms and associated tidal surges, including
Sidr in 2007, Aila in 2009, Mohasen in 2013, Bulbul in 2019 and Amphan in 2020,
causing loss of lives and damages to standing crops and property. It is to be noted
that loss of lives has decreased substantially in recent cyclonic storms. This was
possible due to a paradigm shift in disaster management, from relief and rehabili-
tation to disaster risk reduction. Improved cyclone warning system, fast evacuation
to cyclone shelters and coping strategies of the local communities, which helped in
bringing down loss of lives to a great extent.
According to Hassan 2000, the people of the coast could identify, anticipate and
predict cyclones through the following observations: (a) wind direction; (b) temper-
ature and salinity of sea water; (c) color and shape of the clouds; (d) appearance of
a rainbow; and (e) behavior of certain bird species. Regarding the direction of wind,
the people in the coast believed that a wind blowing from the south-east is more
likely to create a storm, while the wind direction from the north-east has the poten-
tial to generate a cyclone but not a severe one. The wind direction is also associated
with other attributes i.e., a rise in sea water temperature, red-colored cloud and the
appearance of a rainbow if it is day time, suggesting formation of deep depression
in the sea. In most cases, such depressions are formed near the Andaman Islands.
Abnormal behavior of the birds residing in trees is regarded as a signal of a rapid
storm approaching.
People have also developed some survival strategies during and after a cyclonic
event. They are simple tactics like holding onto and binding themselves to trees;
looking for more dependable places like embankments and polders; using floating
items such as timber, thatched roof, straw piles and bunches of coconuts are some
of the survival strategies; these represent spontaneous survival strategies. Generally,
external help and relief goods appear a couple of days after the disaster. During this
intermediary period while they are waiting for relief, people eat stems and roots
of edible plants. For drinking purposes, they often resort to drinking rain water, as
cyclones are always followed by rain for several hours. In the absence of rainwater,
they drink green coconut water. Due to unavailability of medicines, generally, the
victims depend on herbal medicines for treating minor injuries and diarrheal diseases.
As they construct their houses, they use roofing materials and design it in such a
way that the roofs have a sloping toward the wind direction. They plant local varieties
of plants surrounding their homesteads. Haque (2000) states that in Sandwip Island
in the south of Bangladesh, people plant Hurma (Persimmon) trees, which are strong
and can withstand tidal waves. The lives of many people were saved during the April
1991 cyclone, as they held on mangrove trees, like Keora (Sonneratia apetala) and
Sundari (Heritiera fomes). Another interesting phenomenon was that during tidal
surges, people tied rafts to coconut trees so that they rose and fell with the changing
water levels (Haque 2000).
In the north-eastern haor (low depression of wetland) of Bangladesh, a study was
conducted to understand local knowledge and practices regarding early warnings
of a disaster like heavy rainfall, hailstorm or flash floods. Local proverbs such as
“abundance of mango will bring floods and abundance of jackfruit will increase rice
production”, or “abundance of mango and paddy will bring flood” are commonly
used for forecasting early warnings of floods (Islam and Bremer 2016). These “local
scientists”, mostly elderly people interpret early warnings of floods or rainfall by
observing the behavior of animals, birds, insects or amphibians. According to them,
snakes will come to their house before a major flood. Frogs croaking in the month
of March (Bengali month of Chaitra) is a warning for rainfall. If grass hoppers are
flying too high in the middle of April, rainfall will occur. If cattle return to their
home or birds return to their nests, it is sign that a big storm is coming, and if
herons are flying erratically, then this is a particular sign of a northwesterly storm.
Likewise, a persistent southerly wind may be a precursor to heavy rainfall. If water
levels rise coinciding with a cold breeze, a flash flood may be coming. Often senior
citizens are able to provide early warnings based on their experiences and wisdom.
The Haor communities continue to interpret natural signs to predict the weather,
because many say that meteorological weather forecasting systems are incorrect, not
location-specific or inaccessible (Islam and Bremer 2016).
6 Conclusions
The critical role that indigenous knowledge and practices can play in addressing
hazards and improving disaster preparedness is now recognized by disaster risk
reduction specialists. They are yet to be generally used and practiced by communi-
ties, scientists, policymakers and the policy executives. It is believed that local and
indigenous knowledge needs to be integrated with science before it can be used in
policies, education and actions related to disaster risk reduction and climate change.
It is important that such knowledge is promoted to increase the resilience of local
communities against impacts of hazards and climate change. Local and indigenous
152 M. Haque
knowledge is key to increasing resilience of coastal and small island communi-

ties to hydro-meteorological hazards, like tsunami, earthquake and the impacts of
climate change. It is true that indigenous knowledge and practices of the small ethnic
communities have limitations. They mostly remain in oral forms and are not well
documented. There is a move by the academics to validate the knowledge with scien-
tific understanding. Hiwasaki et al. (2014) are of the opinion that a process could be
initiated to identify, document, validate local and indigenous knowledge and inte-
grate this knowledge with science. They termed it as “Local and Indigenous Knowl-
edge and Practices, Inventory, Validation and Establishing Scientific Knowledge
(LIVE)”. Through this method, Hiwasaki et al. (2014) gave scientific explanations
to indigenous knowledge like (a) observations of animal behavior; (b) observations of
celestial bodies; (c) observations of environment; (d) material culture; and (e) tradi-
tional and faith-based beliefs and practices. Hiwasaki et al. observed that the key to
successful implementation of “LIVE Scientific Knowledge” tool is the commitment
of all stakeholders to develop an action plan through a participatory process.
There is no denying that resilience of coastal and small ethnic communities to
meteorological hazards depends on documentation and promotion of their indige-
nous knowledge and practices. Proper validation and scientific explanation of this
knowledge would help the scientists, policymakers and practitioners in developing
an effective country-specific disaster management plan. Following recommendations
of the Hyogo and Sendai Framework for Disaster Risk Reduction, it is suggested that
efforts are to be taken to document and validate the local knowledge and practices
of the island and small ethnic communities of Asia–Pacific, Andaman and Nicobar
and Bangladesh coast to hydro-meteorological hazards, and incorporate them in all
policies, plans and strategies of countries with similar socio-environmental situa-
tions. The 2004 Indian Ocean earthquake and tsunami was a great reawakening,
which called for concerted actions for disaster risk reduction to be undertaken by the
countries and nations bordering the seas.
References
Amin, M.N. 2000. Potential Use of Indigenous Knowledge in Sustainable Conservation of Plant
Diversity in Bangladesh. In Of Popular Wisdom: Indigenous Knowledge and Practices in
Bangladesh, ed. N. A. Khan and S. Sen. Dhaka: Bangladesh Resource Centre for Indigenous
Knowledge.
Chandi, M. 2010. Colonization and Conflict Resolution in the Andaman Islands: Learning from
Reconstruction of Conflict Between Indigenous and Non-indigenous Islanders. In The Jarawa
Tribal Reserve Dossier, Cultural & Biological Diversities in the Andaman Islands, ed. P. Sekhsaria
and V. Pandya. Paris: UNESCO.
Chandi, M., and H. Andrews. 2010. The Jarawa Tribal Reserve: The Last Andaman Forest. In The
Jarawa Tribal Reserve Dossier, Cultural & Biological Diversities in the Andaman Islands, ed. P.
Sekhsaria and V. Pandya. Paris: UNESCO.
Haque, M. 2000. Indigenous Knowledge and Practices in Disaster Management in Bangladesh. In
Of Popular Wisdom: Indigenous Knowledge and Practices in Bangladesh, ed. N.A. Khan and S.
Sen. Dhaka: Bangladesh Resource Centre for Indigenous Knowledge.
Haque, M. 2013. Indigenous Knowledge and Practices in Disaster Management. In Environmental

Governance: Emerging Challenges for Bangladesh. Dhaka: AH Development Publishing House.
Haque, M. 2019. Indigenous Knowledge and Practices in Disaster Management: Experiences of the
Coastal People of Bangladesh. In Disaster Risk Reduction: Community Resilience and Responses,
ed. B. Zutshi, A. Ahmed, and A.B. Srungarapati. Palgrave Macmillan, Springer Nature Singapore.
Hassan, S. 2000. Indigenous Perceptions, Predictions and Survival Strategies Concerning Cyclones
in Bangladesh. In Of Popular Wisdom: Indigenous Knowledge and Practices in Bangladesh, ed.
N.A. Khan and S. Sen. Dhaka: Bangladesh Resource Centre for Indigenous Knowledge.
Hiwasaki, L., E. Luna, Syamsidik, and R. Shaw. 2014. Process for Integrating Local and Indigenous
Knowledge with Science for Hydro-Meteorological Disaster Risk Reduction and Climate Change
Adaptation in Coastal and Small Island Communities. International Journal of Disaster Risk
Reduction 10 (2014): 15–27.
Islam, AKM. S., and S. Bremer. 2016. Signs of the Weather. Dhaka Tribune, 4 June.
McAdoo, B.G., A. Moore, and J. Baumwoll. 2008. Indigenous Knowledge and the Near Field
Population Response During the 2007 Solomon Islands Tsunami. Nat Hazards, 22 January. https://
doi.org/10.1007/s11069-008-9249-z.
Mustafa, M. 2000. Towards an Understanding of Indigenous Knowledge. In Indigenous Knowledge
Development in Bangladesh: Present and Future, ed. P. Sillitoe. Dhaka: The University Press
Ltd.
Pandya, V. 2005. When Land Became Water, Tsunami and the Ongees of Little Andaman Island.
Anthropology News, March.
Rasheed, K.B.S. 2008. Bangladesh, Resources and Environmental Profile. Dhaka: AH Development
Publishing House.
Reddy, S. 2013. Clash of Waves: Post Tsunami Relief and Rehabilitation in Andaman and Nicobar
Islands. New Delhi: Indos Books.
Reddy, S. 2018. New Andamans: The Aftermath of Tsunami in the Andaman and Nicobar Islands.
In The Asian Tsunami and Post Disaster Aid, ed. S. Reddy. Singapore: Springer Nature.
Sekhsaria, P. 2014. The Last Wave, An Island Novel. India: HarperCollins.
Sekhsaria, P. 2017. Islands in Flux, The Andaman and Nicobar Story. India: Harper Litmus.
Shaw, R. 2006. Indian Ocean Tsunami and Aftermath: Need for Environment-Disaster Synergy in
the Reconstruction Process. Disaster Prevention and Management 15 (January, 1).
Sillitoe, P., ed. 2000. Indigenous Knowledge Development in Bangladesh: Present and Future.
Dhaka: The University Press Limited.
Sillitoe, P., P. Dixon, and J. Barr. 1998. IK Research on Floodplains of Bangladesh: The Search for
a Methodology. Grassroots Voice 1 (1): 5–15.
Singh, S.J., M. Fischer-Kowalski, and W. Hass. 2018. The Sustainability of Humanitarian Aid: The
Nicobar Islands as a Case of ‘Complex Disaster’. In The Asian Tsunami and Post Disaster Aid,
ed. S. Reddy. Singapore: Springer Nature.
Walker, D.H., F.L. Sinclair, and R. Muetzelfeldt. 1991. Formal Representation and Use of Indigenous
Ecological Knowledge about Agroforestry Practices: A Pilot Phase Report. School of Agricultural
and Forests Sciences, UK: University of Wales, Bangor.
Warner, K. 1991. Shifting Cultivators: Local Technical Knowledge and Natural Resource Manage-
ment in the Humid Topics. Food and Agriculture Organization of the United Nations.
Driving Factors of Destination Choices
Due to Riverbank Erosion Along
the Brahmaputra River
Sahika Ahmed and Sonia Binte Murshed
Abstract Riverbank erosion is one of the most common natural disasters in

Bangladesh. The after-effect of the flood is one of the major reasons behind this
calamity. Recently, the Brahmaputra riverbank has eroded severely due to frequent
flooding events. As a consequence, the inhabitants of these areas are forced to migrate.
This study was conducted at Dewanganj in Jamalpur district, located at Brahmaputra
River’s left bank. We carried out qualitative research incorporating semi-structured
and unstructured interviews, direct observations, and participatory rural appraisal
(PRA) tools to understand the internal displacement dynamics and the victims’ prob-
lems. Moreover, we divided the study area into five zones, i.e., island char, govern-
ment land, attached char, Gucchogram, upazilla, and town, to understand the spatial
aspects of the displacement pattern. The surveys were done from September 2019
to January 2020 to capture the devastating flood-induced bank erosion impact of the
year 2019. It was noticed that a significant number of people have displaced during
this time frame. This study area’s displacement dynamics are controlled by a total
number of 14 push and pull factors. These push and pull factors show significant
temporal and spatial variations. Among them, the most important are the amount
of resources, land price, land elevation, help from government and NGOs, distance
from ancestors place, etc. The displacement pattern largely depends on these factors.
This study describes an approach to assist planners and policymakers in framing
the current displacement processes and their causes. It will also help plan a better
settlement policy for the affected people, which unfortunately is a considerable part
of the current population.
Keywords Internal displacement · Riverbank erosion · Push factors · Pull factors ·

Flood · Brahmaputra River
S. Ahmed (B) · S. B. Murshed

156 S. Ahmed and S. B. Murshed
1 Introduction
Bangladesh, a riverine country, faces riverbank erosion every year. This forces
millions of its population to displace from their place of origin (Alam et al. 2020; M.F.
Islam and Rashid 1970). Being located at the downstream of the Ganges, Brahma-
putra, and Meghna Rivers, the combined flow (both water and sediment) has a signifi-
cant impact in shaping its hydro-morphology. An extreme amount of sediments from
the Himalayan Mountains is transported by a large monsoon flow to the sea through
the delta of Bangladesh. Consequently, the large rivers behave rather unpredictably
with the permanent risk of riverbank erosion (M. Rahman 2013). Nearly 2.4 billion m.
tons of sediment flows through the rivers of Bangladesh per year (M. Rahman 2013).
This huge amount of washed-out fine soils from the upstream channels are easily
transported downstream and deposited as islands or attached chars. This phenomenon
results in river velocity’s unpredicted flow path at different places and cause erosion
(M. Rahman 2013).
Also, climate change indicates an alteration in temperature and rainfall patterns,
affecting flood magnitude and frequency. Erosion is the after-effect of floods, and
the incidence of abnormal floods is increasing day by day. Environmental impacts,
increased rainfall intensity, infrastructural development activities, and poor drainage
facilities are responsible for increased incidence of abnormal flood in recent years
(Emdad Haque and Zaman 1993). Moreover, as mentioned before, another cause
behind this unpredictable erosion-accretion behavior is the massive amount of
sediments arising from all the river channels of Bangladesh.
Riverbank erosion has catastrophic impacts on the riverine households, which
causes long-term or permanent loss of land (Penning-Rowsell et al. 2013; Haque
and Zaman 1989). People affected by riverbank erosion move for safety and shelter
almost immediately, and then later on move permanently for survival from associated
practical hazards. It harms the social, economic, psychological, and cultural life of
the survivors. Millions of people in Bangladesh are bound to displace or migrate
permanently every year in search of household, earning sources, and cultivable land.
The reasons behind their displacement root in many factors, including education,
health, and better livelihood options. When erosion leads to displacement, people
choose places to move to according to these factors. The erosion victims mainly
displace to other places in the months between July and October. Most of them
face this kind of displacement about 5–7 times in their lives (Afreen 2009). As the
Brahmaputra is predominantly braided and extremely dynamic, erosion and accre-
tion actions are very common (Mondal et al. 2015; Saikia et al. 2019). One hundred
and twenty-seven miles of Brahmaputra, in the lower reach, has eroded almost 30
villages from 2007 to 2017 (Shetu et al. 2017). Even today, erosion is continuing
at the rate of nearly 2000 hectares per year (Alam et al. 2020). Dewanganj is an
Upazila of Jamalpur district which is severely affected by riverine floods and river-
bank erosion every year (Haque et al. 2021). This area is situated at the left bank of
the Brahmaputra River. The unprotected rural area in the Jamalpur district (left bank
of the Brahmaputra River) has a total surface of about 174 km2 and a population
Driving Factors of Destination Choices … 157
of approximately 146,000 people (BBS 2014). As there is no riverbank protection

measures in this rural area, erosion occurs more frequently than on the right bank.
Local victims suffer devastating floods just before the erosion takes place (Mondal
et al. 2015). This area loses a huge amount of land every year due to riverbank erosion.
Inhabitants face loss of households, cultivated lands, and other properties. A signif-
icant number of riverine households face internal displacement multiple times. To
improve this situation, a planned policy on the resettlement of erosion victims will
be helpful if it includes measures for sustainably enhancing the living conditions.
Most research studies on erosion-prone areas (viz. Ashworth et al. 2000; Azam
et al. 2019; Baki and Gan 2012, etc.) dealt with the impact and factors of erosions,
migration, and the geomorphological scenario of the area. Migration policy or how
people manage to settle in a new place is not focused on in most previous studies. A
few studies (viz. Zetter 2012; M.S. Rahman and Gain 2020; M.F. Islam and Rashid
1970) have done some research work on the rehabilitation process. As such, this
study mainly focuses on the factors behind choosing the destination after erosion has
occured.
This study mainly focuses on the sufferings of the erosion victims while choosing
places for habitation after erosion has occured. Specifically, this paper discusses (a)
dominant driving factors behind the selection of displacement destinations and (b)
the spatial variation of push–pull factors.
2 Study Area
The site was selected after observing the severity of flood and erosion in existing
literature and secondary data. The study area is located near the downstream course
of the Brahmaputra, which shifted to this location after the catastrophic flood and
earthquake in 1787. The average channel width is 11 km (Sarker et al. 2003). The
water and sediment discharge of the Brahmaputra River is dominated by the annual
monsoon rainfall (Ashworth et al. 2000), as more than 80% of the river basin’s
yearly precipitation occurs during this season. Bank materials of the Brahmaputra
River consist of loosely packed silt and fine sand with less than 1% clay, so they
are highly susceptible to erosion. The bank erosion rate is as high as 1 km per year
(Klaassen et al. 1993). The mean monthly temperature varies from 12º Celsius in
January to 35.8º Celsius in July. The humidity of this area is high and is the highest
in monsoon between June and October.
The location map of the study area is shown in Fig. 1.
Brahmaputra River is eroding its left bank at Dewanganj Upazila of Jamalpur,
and this area is also prone to riverine monsoon floods. The erosion-accretion process
of the Brahmaputra River develops several small and large sand bars, knows as
“Chars.” Chars that are connected with the mainland are known as “attached chars,”
whereas the chars without any connection to their surrounding area are known as
“island chars.” Our study area includes these two types of chars in addition to three
different locations of the mainland. These are: (i) island char named Halkar Char
Fig. 1 Map of the study area
(Tiner Char), (ii) two attached chars named Kholabari and Chikajani, (iii) two clus-
tered villages named Gujimari Guccho Gram and Adorsho Gram, (iv) two illegally
occupied government land named Badeshshariyabari rail line slum and Chakuriya
rail line slum, and finally, (v) the Dewanganj town. These are low-lying areas at the
floodplain of the Brahmaputra River and located at latitude 25.3 North and longitude
89.76 East.
Charlands in the study area are agriculturally very productive due to their soil
fertility. Also, living there is comparatively cheaper than in other areas. However,
these islands are prone to devastating riverbank erosion and flooding. The condition
of Dewanganj town is quite different as it is only affected by floods. Most of the
people in the islands are farmers. Men typically work in the field when women are
engaging in homestead gardening and animal husbandry. Both men and women work
in the field during harvesting. After a devastating erosion or flooding event, these
people lose their earning source as well as their house. Almost every year, they have
to rebuild their houses somewhere else and find a new income source. Displacements
due to erosion are very common in the study area.
3 Methodology
This research is mainly based on primary qualitative data, where a set of data collec-
tion methods were used. At first, a reconnaissance survey was done to identify the
inhabitants’ current situation and the severity of different natural disasters in recent
years. Though they face consecutive floods and erosions every year, the resettlement
or displacement after being affected by erosion was the primary source of suffering.
Satellite images from Google Earth Pro were also analyzed to understand the extent
of erosion. From the field-level surveys, the factors which influence them in choosing
the destinations were identified.
The reconnaissance survey was done with the help of local government offi-
cials and political persons. However, the following field surveys were done without
involving official and political persons to understand the local perceptions without
getting any biased information. After identifying displacement as their primary
suffering due to erosion, a total of 15 household surveys were done. In each survey,
there were 25–30 persons. These surveys helped to identify the factors that influenced
people when they choose a place for living. Moreover, we followed the open ques-
tionnaire format for our survey, which helped local people express their thoughts
and conceptions without limitations. In these household surveys, both men and
women participated separately. After these surveys, three Focused Group Discus-
sions (FGDs) were done with official persons to confirm some of the factors behind
choosing migration destinations. These FGDs were done with local primary teachers,
local political persons (members, chairman), and local government authorities (UNO
or subdistrict executive officer & TNO thana executive officer). These FGDs and
household surveys were the main primary sources of collected data.
During the second field visit, the study area was divided into five categories based
on the five different land types. These are: island char, attached char, illegally occu-
pied government land, cluster village (Gucchogram), and Dewanganj town. After
that, ten FGDs were conducted with the local groups of those selected five places.
From our first visit, we have identified a total of 14 factors that helped in choosing
migration destinations. So, prioritizing these factors was the main focus of the FGDs
of the second visit. These FGDs were done with structured questionnaires where the
factors were classified into two groups (push factors and pull factors), and then, the
evaluation of these factors was done. The dominant driving factors were then ranked
by structured questionnaire surveys. The questionnaire surveys were based on the
Fig. 2 Methodology of this study
preference ranking tool of PRA. The final ranking was achieved after analyzing all
the outputs from FGDs and questionnaire surveys (Fig. 2).
4 Results and Discussions
The erosion that occurred in 2019 had nearly destroyed a whole union named
Kholabari with government buildings, pakka (or concreted) houses, schools,
mosques, and concrete roads. Many government infrastructures eroded in Dewan-
ganj. Many of the inhabitants shared the incident of the Harindhara embankment,
which eroded ten years ago. From then onward, this area does not have any concrete
embankment. As such, people are suffering from more severe erosion now, and the
recent devastating flood of 2019 broke the past records of severity.
Due to erosion, the inhabitants are forced to settle in different places every year.
Choosing a destination for migration is not easy as there are lots of uncertainties
involved in this process. The extent of erosion is also unpredictable. From Fig. 3, the
Landsat satellite image of 2020 shows that the distance of the left riverbank from
the railway station was 3.74 km, whereas this distance was 5.63 km in 1998. So, this
Fig. 3 A comparison of satellite images of the Brahmaputra riverbank at Dewanganj between a

1998 and b 2020
extreme riverbank erosion destroyed many people’s land and belongings, along with
government infrastructures.
A total of 15 factors were identified (Table 1) as push and pull factors from the
FGDs, household interviews, and Key Informant Interviews (KIIs). Push factors
force people to move away from their inhabitation, whereas pull factors motivate
people to move into a new location for different socioeconomic reasons. These 15
factors act differently for people of different zones. The factors are given in Table 1.
Table 1 Factors identified

Study zone Push factors Pull factors
from FCDs
Island char Erosion Surplus of resources
Lower income Daily expenses
facilities Pastureland
Lower land elevation Relative’s Land
Marital status Land cost
Distance from origin Link with political
Poor education persons
facilities Government help or
Inferior living NGO’s help
condition
Manpower
Attached char Lower income Government help or
facilities NGO’s help
Erosion Relative’s Land
Daily expenses Better education
Lower land elevation facilities
Distance from origin Land Cost
Scarcity of resources Pastureland
Manpower Better living
Link with political condition
persons Link with political
Marital status persons
Government land Lower income Government help or
facilities NGO’s help
Manpower Link with political
Distance from origin persons
Erosion Land cost
Relative’s land Higher land
Marital status elevation
Scarcity of resources Better education
Inferior living facilities
condition
Barren land
Daily expenses
Gucchogram Lower land elevation Land cost
Marital status Government help or
Erosion NGO’s help
Relative’s land Link with political
Barren land persons
Lower income Better living
facilities condition
Manpower Surplus of resources
Poor education
facilities
Distance from origin
Daily expenses
(continued)
Table 1 (continued)
Study zone Push factors Pull factors
Dewanganj town Erosion Better living
Lower income condition
facilities Distance from origin
Poor education Higher land
facilities elevation
Land cost
Daily expenses
Barren land
Scarcity of resources
Another key focus of our study is to understand the spatial variation among the
push and pull factors. The preference ranking tool helped to prioritize the 15 key
factors that have been identified. Tables 2 and 3 show the preference ranking of push
and pull factors, respectively.
The above results show the local inhabitants’ varying acceptance levels when it
comes to different push and pull factors. Our findings also depict that the major push
factors are almost common in these areas.
Due to riverbank erosions, the inhabitants of these areas become vulnerable from
different perspectives; for example, they do not have any long-term plans or secu-
rity of education, income facilities, and health. After being homeless, they face
various inconveniences every day. In addition to riverbank erosion, the unstable and
unplanned condition of their lives creates other reasons for their displacement. This
study shows that the other factors excluding erosion can sometimes become the
main reason for displacement of people in a particular area. The different zones have
different types of benefits. Local people usually take benefits from the surrounding
environment and have become accustomed to it. That is a major reason for people’s
non-migratory behavior. They prefer to displace to places nearer to their origins.
Push factors were identified as factors which make them compelled or interested in
going for new habitation. Displaced households were interviewed to understand the
local push factors. The ranking tables (Tables 2 and 3) show that erosion, one of the
topmost push factors, is common across the five different zones. These devastating
phenomena compelled them to find new accommodations. Though it is supposed
to be ranked top among all the factors, it is noticeable that the need for income
facilities remains unbeatable. With the development of education and living style,
more and more people are searching for earning sources beyond their living places.
Besides, the opportunity for different types of jobs is rare in these locations. Most
of the people earn their livelihood by cultivation, fishing, and temporary work on
government projects. There are some facilities for academic work in Dewanganj
town, but that is not enough for all the educated and skilled persons of these areas.
Also, women do not have many facilities over there. For survival, many women are
forced to move into city areas to work in the garments sectors, domestic sectors, etc.
So, lack of income facilities is found to be one of the primary reasons for permanent
migration.
Table 2 Ranking of push factors across five zones of Dewanganj Upazila

Ranking Island char A ached char GOVT land Gucchogram Dewanganj town
of push
factors of
different
places
1 Erosion Lower income Lower Lower land Erosion

facili es income Eleva on
facili es
2 Lower income Erosion Manpower Marital Status Lower income
facili es facili es
3 Lower land Daily expenses Distance Erosion Poor educa on
eleva on from Origin facili es
4 Marital Status Lower land Erosion Rela ve's Land Land cost
eleva on
5 Distance from Distance from Rela ve's Barren land Daily expenses
origin Origin Land
6 Poor educa on Scarcity of Marital Lower income Barren land

facili es resources Status facili es
7 Inferior living Manpower Scarcity of Manpower Scarcity of resources

condi on resources
8 Manpower Link with Inferior Poor educa on N/A

poli cal living facili es
persons condi on
9 N/A Marital Status Barren land Distance from N/A

Origin
10 N/A N/A Daily Daily expenses N/A

expenses
Due to the occurrence of devastating floods every year, land elevation is also
an essential factor here. Naturally, island chars are not permanent. These locations
face erosion and deposition more frequently than any other zones. Subduction of
soil is also very common. So, these people who live in less elevated land suffer
the most during flooding events. Hence, elevated land is very desirable and often
gets prioritized as migration destinations. Also, local inhabitants in gucchograms
and attached chars do not lose any chance to migrate or displace to higher elevated
locations.
In addition, the erosion victims want to migrate to safer locations immediately
after losing their homesteads. In those cases, the distance of migrated destinations is
not an issue. Nevertheless, they always plan to return to their origin, near their lost
household, after being settled. However, sometimes, the opposite case happens. If
the number of adults and earning members is high in a family, they usually migrate
to cities and do not plan to return.
Table 3 Ranking of pull factors across five zones of Dewanganj Upazila
Ranking of pull Island char A ached GOVT land Gucchogram Dewanganj

factors of char town
different places
1 Surplus of Government Government Land Cost Living

resources help or help or Condi on
NGO's help NGO's help
2 Daily Rela ve's Link with Government Distance

expenses Land poli cal help or from Origin
persons NGO's help
3 Pastureland Be er Land Cost Link with Higher land

educa on poli cal Eleva on
facili es persons
4 Rela ve's Land Cost Higher land Be er living N/A

Land Eleva on condi on
5 Land Cost Pastureland Educa on Surplus of N/A

resources
6 Link with Be er living N/A N/A N/A

poli cal condi on
persons
7 Government Link with N/A N/A N/A

help or poli cal
NGO's help persons
Usually, families want to settle their daughters (after marriage) at distant places
from the origin to keep them free from the suffering of erosion. Also, in recent days,
male and female inhabitants, who are not unmarried, move to towns by themselves
and stay there permanently by marrying someone and settling there.
The facility of education and health service is very inferior in island chars. In
island char areas, very few families consider poor education opportunities as the
push factors for choosing their migration destinations. The exception is Dewanganj
town, where the migration rate due to education facilities is relatively higher.
In the island chars, the daily expense is very low as the amount of natural resources
is very high. They do not have to pay much for their food and fuels as the charland
areas are productive enough to supply these demands at a meager cost or no cost at all.
Besides, the living standard of the local inhabitants of island char is very low. There-
fore, many families with lower income facilities prefer island char when displacing
after erosion. Getting help from relatives is another critical factor in choosing new
destinations. A significant number of people chose to shift close to relatives immedi-

ately after displacement. Moreover, if anyone has any professional or family link with
any political persons, it is easier to get land in island char. The amount of cultivable
land and pasture is abundant in island chars, which is a pull factor for the people
with lower income from the other three zones, excluding Dewanganj town.
The vast amount of natural resources is the main pull factor of island chars which
influences them not to choose other locations except another island char. Also, the
land cost is relatively very low here. However, the reliefs from government or NGOs
seldom reach the island chars. The people in attached chars, goverment lands, and
gucchograms are more likely to benefit from relief services. Hence, access to relief
services works as a pull factor for these places.
The displaced people who are illegally residing in goverment lands seek help from
political persons, as they are afraid of eviction. That is why a link with a political
persons is a pull factor for the area. Elevated land is also a pull factor here as it saves
them from the severity of floods. The education system and the developed medical
facilities are the pull factors for all the areas, excluding island char. In gucchograms,
the land is given by the government of Bangladesh and is free of cost. That is why
people living here do not want to displace easily.
Though these ranking positions seem different in different seasons according to
their existing situation (such as during the season of riverbank erosion it is the main
and top-ranking push factor, but in the dry season they may focus on income facilities
or other factors), this study has focused on the overall sufferings and preferences of
the inhabitants. This study is done in both wet and dry seasons and the final output
is the merged conclusion of their overall condition.
These two tables (Tables 1 and 2) introduce the ranking of the factors. Neverthe-
less, if we can understand the reason behind these rankings, it will be easy to identify
the needs and sufferings of these people. It will help to design and plan for better
habitation for them. To develop the condition of these people, a better adaptation
strategy to natural disasters is essential. Moreover, a further study on the driving
factors will also be helpful in this matter.
5 Conclusion
Natural disasters cannot be avoided or stopped as these are beyond the control of
humankind. Poor people who are continuing to lose their properties due to these
disasters remain poor. Land losses make their condition even more vulnerable; it
hampers the lifestyle of an entire community. Although we cannot control these
situations, the implication of some planned projects to help victims return to their
usual lifestyle will make a huge difference in the ultimate development of a society.
In Bangladesh, people who are living by the river are facing issues of riverbank
erosion every year. So displacement of habitation is common after these incidents.
People living in different zones have different choices and demands, which should
be kept in mind when planning any policy for rehabilitating these people.
The government of Bangladesh has already taken projects, including Gucchogram

(cluster village), under the Adarsha Gram Project. Although the project was planned
for people affected by cyclones, many cluster villages are also made in the erosion-
prone areas in recent years. Ranking the push and pull factors behind their choice of
rehabilitation place would be helpful to achieve sustainability of such projects.
The output of this case study has shown the difference of choices in different
zones by ranking these factors. Also, with the help of the push factors’ ranking, the
importance of common factors like governmental help or NGO’s help, land cost,
and link with political persons is easily noticeable. The factors known as pull factors
(such as high land elevation, better income facilities, surplus of natural resources,
better education facilities, protective measures against flood and erosion) should be
introduced and ensured in the zones where necessary. In contrast, the factors known
as push factors (such as poor income facilities, inferior living conditions, scarcity of
resources) should be eliminated or adapted in feasible and secure ways.
Acknowledgements This study was funded by the South Asia Consortium for Interdisciplinary
Water Resources Studies (SaciWATERs) as part of their South Asian Water (SAWA) Leadership
program grant. The authors are also grateful to the local people of Dewanganj Upazila. Special
thanks to Sumaiya Binte Islam, Khadizatul kubra, and Md. Shadhin Hossain for helping to gather
field information during the Covid era.
Appendix
See Figs. 4 and 5.

Fig. 4 Ranking of push factors for different zones a Island char, b Attached char, c Government
land, d Gucchogram, and e Dewanganj Town
Fig. 5 Ranking of pull factors for different zones a Island char, b Attached char, c Government
land, d Gucchogram, and e Dewanganj Town
References
Afreen, Farzana. 2009. Study on Bank Erosion and Its Impact on the Livelihood of the Local
People: A Case Study along the Nabaganga River. M.Sc in WRD, Institute of Water and Flood
Management, Bangladesh University of Engineering and Technology.
Alam, G.M. Monirul, Khorshed Alam, Shahbaz Mushtaq, Md Nazirul Islam Sarker, and Moazzem
Hossain. 2020. Hazards, Food Insecurity and Human Displacement in Rural Riverine Bangladesh:
Implications for Policy. International Journal of Disaster Risk Reduction 43. https://doi.org/10.
1016/j.ijdrr.2019.101364.
Ashworth, Philip J., James L. Best, Julie E. Roden, Charles S. Bristow, and Gerrit J. Klaassen. 2000.
Morphological Evolution and Dynamics of a Large, Sand Braid-Bar, Jamuna River, Bangladesh.
Sedimentology. https://doi.org/10.1046/j.1365-3091.2000.00305.x.
Azam, Gausul et al. 2019. Climate Change and Natural Hazards Vulnerability of Char Land (Bar
Land) Communities of Bangladesh: Application of the Livelihood Vulnerability Index (LVI).
Global Social Welfare, Springer.
Baki, Abul Basar M., and Thian Yew Gan. 2012. Riverbank Migration and Island Dynamics of the
Braided Jamuna River of the Ganges-Brahmaputra Basin Using Multi-Temporal Landsat Images.
Quaternary International 263: 148–61.
BBS. 2014. Population & Housing Census-2011. Bangladesh Bureau of Statistics.
Emdad Haque, C., and M.Q. Zaman. 1993. Human Responses to Riverine Hazards in Bangladesh:
A Proposal for Sustainable Floodplain Development. World Development 21 (1): 93–107. https://
doi.org/10.1016/0305-750X(93)90139-Z.
Haque, C.E., and M.Q. Zaman. 1989. Coping with Riverbank Erosion Hazard and Displacement in
Bangladesh: Survival Strategies and Adjustments. Disasters 13 (4): 300–314. https://doi.org/10.
1111/j.1467-7717.1989.tb00724.x.
Haque, M.M., Islam, S., Sikder, M.B., and Islam, M.S. 2021. Assessment of Flood Vulnerability in
Jamuna Floodplain: A Case Study at Jamalpur District, Bangladesh. Research Square, 1–18.
Islam, M.D. Fakrul, and A.N.M. Baslur Rashid. 1970. Riverbank Erosion Displacees in Bangladesh:
Need for Institutional Response and Policy Intervention. Bangladesh Journal of Bioethics. https://
doi.org/10.3329/bioethics.v2i2.9540.
Klaassen, G.J., E. Mosselman, and H. Brühl. 1993. On the Prediction of Planform Changes in
Braided Sand-Bed Rivers. Advances in Hydro-Science and -Engineering.
Mondal, M. Shahjahan, M. Aminur Rahman, Nandan Mukherjee, Hamidul Huq, and Rezaur
Rahman. 2015. Hydro-Climatic Hazards for Crops and Cropping System in the Chars of the
Jamuna River and Potential Adaptation Options. Natural Hazards 76 (3): 1431–1455. https://doi.
org/10.1007/s11069-014-1424-9.
Penning-Rowsell, Edmund C., Parvin Sultana, and Paul M. Thompson. 2013. The ‘Last Resort’?
Population Movement in Response to Climate-Related Hazards in Bangladesh. Environmental
Science and Policy 27: S44-59. https://doi.org/10.1016/j.envsci.2012.03.009.
Rahman, Md Sadequr, and Animesh Gain. 2020. Adaptation to River Bank Erosion Induced
Displacement in Koyra Upazila of Bangladesh. Progress in Disaster Science 5. https://doi.org/
10.1016/j.pdisas.2019.100055.
Rahman, M.R. 2013. Impact of Riverbank Erosion Hazard in the Jamuna Floodplain Areas in
Bangladesh. Journal of Science Foundation 8 (1–2): 55–65. https://doi.org/10.3329/jsf.v8i1-2.
14627.
Saikia, Lalit, Chandan Mahanta, Abhijit Mukherjee, and Suranjana Bhaswati Borah. 2019. Erosion–
Deposition and Land Use/land Cover of the Brahmaputra River in Assam, India. Journal of Earth
System Science 128 (8). https://doi.org/10.1007/s12040-019-1233-3.
Sarker, Maminul Haque, Rob Koudstaal, and Mustafa Alam. 2003. Rivers, Chars and Char Dwellers
of Bangladesh. International Journal of River Basin Management. https://doi.org/10.1080/157
15124.2003.9635193.
Shetu, M.S.R., M.A. Islam, K.M.M. Rahman, and M. Anisuzzaman. 2017. Population Displacement
due to River Erosion in Sirajganj District: Impact on Food Security and Socio-Economic Status.
Journal of the Bangladesh Agricultural University 14 (2): 191–99.
Zetter, R. 2012. Shelter and Settlement for Forcibly Displaced People. In International Encyclopedia
of Housing and Home, 330–335. https://doi.org/10.1016/B978-0-08-047163-1.00024-2.
Bivariate Drought Risk Estimation Using
a Multivariate Standardized Drought
Index in Marathwada Region, India
Rajarshi Datta and Manne Janga Reddy
Abstract Drought is a natural disaster that can cause water scarcity and damage to
crop yields. Rather than conventionally used univariate drought monitoring indices,
this study applied the parametric Multivariate Standardized Drought Index (MSDI)
for drought monitoring in the Marathwada region. The index is constructed using
historical time series of precipitation and soil moisture by engaging copula func-
tions. The drought conditions characterized by MSDI are then compared with two
univariate drought indices. Two significant drought characteristics, duration, and
severity are identified using the MSDI and fitted probability models. The best-fit
marginal distributions were selected by performing goodness-of-fit tests and stan-
dard performance measures. Three Archimedean copulas and two meta-elliptical
copulas were applied for bivariate modelling of drought characteristics, and their
suitability was evaluated using goodness-of-fit tests. Subsequently, the drought risks
for the study region have been assessed using the constructed copula-based joint
distribution models. The results highlight the importance of multivariate drought
risk assessment in the study region.
Keywords Drought · Copula · Risk · Precipitation · Soil moisture · Joint

distribution
1 Introduction
Droughts are natural disasters that have detrimental effects on various sectors. The
concept of drought and different types of droughts have been widely reviewed (Mishra
R. Datta (B) · M. J. Reddy

Department of Civil Engineering, Indian Institute of Technology Bombay, Mumbai, India
M. J. Reddy
M. J. Reddy
Interdisciplinary Programme in Climate Studies, Indian Institute of Technology Bombay,
Mumbai, India
174 Rajarshi et al.
and Singh 2010). Drought events and their impacts are often region-specific because
of regional differences in hydro-meteorological variables and socioeconomic factors.
In general, meteorological drought can be defined as a prolonged deficiency in rainfall
in a particular area, compared to the statistical multi-year mean. Drought is triggered
by anthropogenic activities such as deforestation and groundwater extraction, as well
as climatic variability (i.e. temperature and precipitation anomaly). These activities
can impact the local water balance and alter the propagation of droughts (Liu et al.
2016). Moreover, global warming caused by human activities can make drought
events worse by intensifying their severity and duration (Ahmadalipour et al. 2017).
The meteorological, hydrological, agricultural, and socioeconomic drought types
are the four main drought classifications. Meteorological drought is defined as precip-
itation deficiency over a region for an extended period. When meteorological drought
is prolonged, the surface and sub-surface water levels deplete, and it leads to hydro-
logical drought. If the soil moisture and rainfall are inadequate to sustain the crop
growth, then eventually, the crop starts wilting, resulting in agricultural drought.
Socioeconomic drought begins when existing resources become further incapable of
meeting the growing water demands (Liu et al. 2020).
Drought indices are numerical expressions based on climatic and hydrological
variables, which are used for assessing drought’s impact and defining drought char-
acteristics. In the past decades, various drought indices have been introduced, and
among them, the Palmer Drought Severity Index (PDSI) and the Standardized Precip-
itation Index (SPI) are commonly applied for drought monitoring (Mishra and Singh
2010). The PDSI is based on a physical water balance model (Palmer 1965). McKee
et al. (1993) derived the SPI, which utilizes long-term rainfall records.
The Standardized Soil Moisture Index (SSI) (Hao and AghaKouchak 2013) is
similar to the SPI, which uses soil moisture as its input parameter. Hao and AghaK-
ouchak (2013) presented the Multivariate Standardized Drought Index (MSDI)
by combining the rainfall and soil moisture data, which can be applied for
agro-meteorological drought monitoring.
The copula functions developed by Sklar (1959) are handy functions that can be
useful for capturing joint dependency between the random variables, irrespective of
their marginal distribution. Hence, several studies (Janga Reddy and Ganguli 2012;
Ganguli and Reddy 2012; Mirabbasi et al. 2012) applied a copula-based framework
for drought frequency analysis.
Recent reviews (Hao and Singh 2015) emphasize the monitoring of drought from
a multivariate perspective. Motivated by the above facts, the study takes up two main
objectives (1) to compare the historical drought events based on MSDI, SPI, and SSI;
and (2) to apply copula-based modelling to assess the agro-meteorological drought
risks based on 3-month MSDI in the study region.
The rest of the paper is arranged as follows: The study area, dataset, and method-
ology are described in Sect. 2. The application, results, and discussion are presented
in Sect. 3. Section 4 presents the conclusions of the study.
Bivariate Drought Risk Estimation Using … 175
2 Materials and Methodology
2.1 Study Area and Dataset
The meteorological subdivision of Marathwada is situated in the central and south-

eastern parts of Maharashtra state. The geographical extent of latitudes of the region
is 17°35’N–20°40’N, and the expansion of longitudes is 74°40’E–78°15’E as shown
in (Fig. 1). The entire region is divided into eight districts, covering an area of
64,590 km2 . The annual rainfall over Marathwada is 826 mm, more than 80% of
which is dependent on the southwest monsoon. A tropical semi-arid climate prevails
in the region, and the farmers in the area experience severe distress. The highest
temperature (45 °C) is observed during May, and the coldest temperature (11 °C)
occurs in December.
This study used the rainfall data of 91 grid points over the Marathwada region
prepared by the India Meteorological Department (IMD), Pune (Pai et al. 2014).
This data has a spatial resolution of 0.25° × 0.25°, and we extracted the data
for all the grid points from 1980–2019. The Thiessen polygon approach has been
adopted to compute the average areal precipitation over the Marathwada Region.
The Thiessen polygons depicted the representation area of each grid, which has been
used in the computation of mean precipitation. The monthly root zone soil moisture
(RZSM) was downloaded from the Modern-Era Retrospective analysis for Research
Fig. 1 Location map of the study region

176 Rajarshi et al.
and Applications 2 (MERRA-2) (Gelaro et al. 2017; Reichle et al. 2017) for the
duration 1980–2019 from (https://gmao.gsfc.nasa.gov/reanalysis/MERRA-2/data_a
ccess/). The spatial resolution of the dataset is 0.5° latitude × 0.625° longitude, which
is resampled to 0.25° latitude × 0.25° longitude by bilinear interpolation. The mean
moisture content over 0–100 cm depth of soil is considered as the root zone soil
moisture (m3 /m3 ). Sathyanadh et al. (2016) assessed different soil moisture (SM)
datasets against India Meteorological Department’s in-situ measurements spread
over India. The MERRA SM product exhibited a good correlation with the in-situ
measurements. Several drought-related studies worldwide have used the MERRA
soil moisture product (Farahmand and AghaKouchak 2015; Agutu et al. 2020; Chen
et al. 2020).
2.2 Methodology
2.2.1 Formation of Standardized Indices
McKee et al. (1993) recommended two-parameter gamma distribution for obtaining

the cumulative distribution function (CDF) from the observed rainfall data. The
gamma probability density function (PDF) is expressed as:
1 χ
g(χ ) = χ α−1 e− β (1)
β α (α)
where (α) represents the gamma function, α and β stand for the shape and rate
parameters respectively, and χ is a random variable that accepts non-negative and
nonzero values. By taking t = χβ̂ , the cumulative probability G(χ ) can be expressed
as:
χ
1
G(χ ) = t α̂−1 e−t dt (2)
α̂
0
Equation 2 is not applicable for zero rainfall values. The modified CDF can be
written as:
H (χ ) = q + (1 − q)G(χ ) (3)
In the above equation, q is the probability of zero precipitation. The CDF of

the mixed distribution can then be transformed to the CDF of the standard normal
distribution to obtain the standardized precipitation index.
Unlike SPI, there is no specific recommendation regarding best-fit distribution for
soil moisture. Farahmand and AghaKouchak (2015) used a non-parametric approach
by fitting soil moisture data to an empirical distribution. The SSI values in their study
were in the range of −2 to + 2, whereas it is observed that, in general, the values

of the standardized indices lie between −3 and + 3. According to (Soľáková et al.
2014), the fact that parametric distribution captures extreme values better than their
non-parametric counterparts can partly explain this result. Zhang et al. (2017) applied
the gamma distribution to characterize SPI and SSI over the four Northern Indian
states of the Indo-Gangetic plain. Following this work, we have used the gamma
distribution to compute SSI over the Marathwada region.
2.2.2 Copula
Sklar (1959) first introduced the copula functions. One-dimensional marginal distri-
bution functions can be joined or ‘coupled’ to construct multivariate distribution
functions using copulas (Nelsen 2007). The copula functions can exhibit the depen-
dence pattern among multiple random variables, and they are considered an efficient
tool for joint distribution modelling (Joe 1997; Nelsen 2007).
Sklar’s theorem gives the explicit definition of the copula, which is restated below.
If x1 , x2 , . . . . . . xn are n random variables, and their corresponding marginal distri-
bution functions are given as G 1 (x1 ), G 2 (x2 ) . . . . . . .., G n (xn ), then the copula func-
tion C can merge those univariate distributions to produce their joint cumulative
distribution:
G(x1 , x2 , . . . ., xn ) = C{G 1 (x1 ), G 2 (x2 ) . . . . . . .., G n (xn )}

= C(u 1 , .., u n ), ∀x1 , x2 , . . . xn ∈ R (4)
The study used three Archimedean and two elliptical copulas for the multivariate
modelling. The functional form of the different copulas and their parameter ranges
are presented in Table 1. (Zhang and Singh 2019) can be referred for more details
regarding the copula functions.
2.2.3 Estimation of the Copula Parameters
The maximum pseudo-likelihood (MPL) technique has been adopted for deriving the
parameters. In this method, instead of parametric distributions, the copula parameters
are estimated using rank-based empirical marginal distributions. The maximization
of the pseudo-log-likelihood
function gives the required parameters. Given a random
vector X ∈ X i,1 , ....., X i,d , the pseudo-likelihood can be computed as follows:

n

Lθ = ln cθ Ui,1 , ...., Ui,d (5)
i=1
Ri,d
where Ui,d = n+1
,d = 2 and Ri,d is the ranked data of X i,d .
178
Table 1 List of the copula families

Copula class Type Copula CDF Cθ (u, v) Parameters º
−1
Archimedean Clayton max [u −θ + v −θ − 1 θ , 0) θ ∈ [−1, ∞)\{0}
1
Gumbel-Hougaard exp{−[(−lnu)θ + (−lnv)θ ] θ } θ ∈ [1, ∞)

−θu −θv
e −1 e −1
Frank − θ1 ln 1 + θ ∈ (−∞, ∞)\{0}
(e−θ −1)

Elliptical Gaussian R −1 (u), −1 (v), ρ −1 ≤ ρ ≤ 1
− ν+2
tν−1 (u) tν−1 (v) 2
1 x 2 − 2r x y + y 2
√ 1+
2
dxdy
−∞ −∞ 2
1 − r 2 ν 1−r
Student’s t ν > 2, r ∈ (0, 1]
− ν+1
2
x ν+1 y2
tν (x) = √ 2ν 1 + dy
−∞
ν 2 ν
Rajarshi et al.
Then we can estimate the required parameter as:
θ̂ = arg max{L θ } (6)
2.2.4 Selection of Best Copula
In general, the copula function with the highest likelihood can be considered as the
best-fit copula. However, goodness-of-fit tests are required to ensure the suitability of
the copula model. These tests are formulated by considering the distance between the
empirical and theoretical copulas. The Cramer-von Mises (CvM), Anderson–Darling
(AD), and Integrated Anderson–Darling (IAD) test statistics have been computed to
decide the best performing copula model.
The empirical copula function is expressed as (Genest and Favre 2007):

1
n
Ri,1 Ri,2
Cn = I ≤ u1, ≤ u2
n i=1 n+1 n+1 (7)
u = (u 1 , u 2 ) ∈ [0, 1] 2
where u 1 and u 2 are random variables and the indicator function of set A is denoted
by I.
The CvM test statistic can be expressed as:

n
2
Sn = Cn Ui,1 , Ui,2 − Cθ Ui,1 , Ui,2 (8)
i=1
where Cn and Cθ are empirical and theoretical copula, respectively.

The Anderson–Darling and Integrated Anderson–Darling statistics can be
expressed as (Ane and Kharoubi 2003):
|Cn (i/n, j/n) − Cθ (i/n, j/n)|

AD = max √ (9)
1≤i≤n,1≤ j≤n Cθ (i/n, j/n)[1 − Cθ (i/n, j/n)]
n n
[Cn (i/n, j/n) − Cθ (i/n, j/n)]2
I AD = (10)
C (i/n, j/n)[1 − Cθ (i/n, j/n)]
i=1 j=1 θ
In the above equations, the order statistics of random variables are represented by
i and j.
180 Rajarshi et al.
2.2.5 The Formulation of the Multivariate Standardized Drought

Index (MSDI)
The parametric MSDI (Hao and AghaKouchak 2013) is adopted for drought charac-
terization in the study. The MSDI integrates the SPI and SSI and can be expressed
as:
M S D I = −1 ( p) (11)
where p is the joint cumulative probability of the joint distribution of precipitation

and soil moisture.
p = P(U < u, V < v) = C[FU (u), FV (v)] (12)
where C is the copula function, and FU (u), FV (v) are the CDFs of the random
variable U and V, respectively. The joint CDF of precipitation and soil mois-
ture is inversely transformed by the standard normal distribution, resulting in the
construction of MSDI, as given in Eq. 11.
2.2.6 Drought Identification and Characterization
The concept of run theory (Yevjevich 1967) is applied for computing two drought
characteristics—duration and severity. Based on the theory, a threshold is defined.
We considered a threshold value of −0.8 (Hao and AghaKouchak 2013). If the index
falls below the threshold, a drought event is identified, and the event terminates if
the index value is greater than the threshold value. The consecutive time interval
(in months) between the onset and termination of a drought event is known as the
duration (D), and severity (S) is the summation of the index value over the time
interval. The maximum absolute value of the index over the duration is known as
drought peak (P). The drought inter-arrival time (L) is the consecutive number of
months counting from the starting point of a drought episode to the onset of the next
episode (Song and Singh 2010) (Fig. 2).
2.2.7 Return Period Analysis of Droughts
The computation of the return period is necessary for analysing drought risk. In this
study, we computed two different cases of bivariate return periods and inter-compared
them with univariate drought frequency. The univariate return periods for duration
and severity can be separately expressed using the following equations (Willems
2000; Kim and Valdés 2003):
E(Id )
Td = (13)
1 − FD (d)
Fig. 2 Illustration of the drought events. Note D is Duration, P is Peak, S is Severity, L is Inter
arrival time
E(Id )
Ts = (14)
1 − FS (s)
In the above equations, Td = Return period considering duration is greater than

or equal to a particular value, and Ts = Return period considering severity is greater
than or equal to a specific value. E(Id ) is the expected inter-arrival time, FD (d) and
FS (s) are the CDF of duration and severity, respectively.
However, when the underlying random variables are mutually correlated, the
multivariate return period plays a more significant role in the design criteria. The
drought events are considered bivariate, and hence the return period should be
computed while considering the joint behaviour of the drought characteristics. The
joint return period can be evaluated for two separate cases (Shiau 2003): the ‘OR’
case in which duration or severity is exceeding or equal to a specific magnitude, and
the ‘AND’ case where the duration and severity is exceeding or equal to a particular
OR AN D
magnitude, which is designated by TDS and TDS respectively as:
E(Id )
AN D
TDS = (15)
1 − FD (d) − FS (s) + C(FD (d), FS (s))
E(Id )
OR
TDS = (16)
1 − C(FD (d), FS (s))
where C is the selected copula function.

182 Rajarshi et al.
Table 2 Copula parameters and corresponding test statistics for modelling of MSDI
Copula Functions Parameter Log-likelihood AIC Sn AD IAD
Value
Clayton 1.0145 85.477 −168.9542 0.8818 0.2098 4.5726
Frank 5.9684 161.4117 −320.8235 0.1328 0.1903 1.0652
Gumbel-Hougaard 1.8244 135.4227 −268.8454 0.2807 0.1912 1.7045
Gaussian 0.6729 144.6373 −287.2746 0.2580 0.1893 1.6540
3.1 Selection of Best Copula for MSDI Construction
The present study investigated agro-meteorological droughts considering the inputs

of rainfall and soil moisture and performed drought analysis using SPI, SSI, and
MSDI. Four copulas were evaluated to measure the mutual dependence between
precipitation and soil moisture. The estimated copula parameters using the MPL
approach and the associated goodness-of-fit test statistics are given in Table 2. The
Frank copula is further selected to construct MSDI at a 3-month time scale (MSDI-3)
for the study region.
3.2 Performance Assessment of MSDI
The MSDI has been compared with SPI and SSI at a 3-month time scale (SPI-3 and
SSI-3), and the plot is depicted in Fig. 3. The MSDI may not give the same duration
and severity as other univariate indices, because the joint distribution at any given
quantile of the univariate distribution may not be the same. As seen from the previous
analysis, Frank copula has been selected for constructing the joint index. The initial
hypothesis is that the copula-based MSDI would perform better than SPI and SSI.
During the entire study period of 1981–2019, SPI and SSI, in many cases, mani-
fested substantial differences within severity and duration. If we consider a period,
for instance, 2008–2013, it is observed that the onset of drought is exhibited by the
SPI before the SSI.
The inter-comparison indicates that MSDI could capture the severity of drought
similar to SSI, and drought onset identical to the SPI. It was also observed that when
both the univariate indices indicate drought, MSDI also showed drought. The MSDI
index was further applied for bivariate drought risk estimation in the Marathwada
region.
A total of 48 drought events were recognized from the estimated MSDI-3 series.
Summary statistics of drought characteristics for the study period of 1981–2019 are
presented in Table 3.
Fig. 3 Parametric drought indices at 3-month time scale for the Marathwada region
Table 3 Summary statistics of droughts based on MSDI, SSI, and SPI in the Marathwada region
Climate variables Statistics Drought Index
MSDI SSI SPI
Drought Number of droughts 48 29 41
Average inter-arrival time (months) 9.75 16 11.32
Duration (months) Mean 3.85 3.55 2.29
Maximum 12 12 8
Standard deviation 3.05 3.07 1.62
Skewness 1.22 1.54 1.42
Kurtosis 3.99 4.78 5.15
Drought severity Mean 5.54 3.99 3.18
Maximum 20.97 14.71 15.09
Standard deviation 5.42 3.75 2.84
Skewness 1.55 1.42 2.17
Kurtosis 4.67 4.24 8.85
Pearson’s correlation coefficient, along with Kendall’s tau and Spearman’s rho,
have been computed to check the dependence between two drought variables. Table
4 illustrates the correlation coefficient and associated p-values. The results indicated
a significant correlation between drought characteristics.
184 Rajarshi et al.
Table 4 Dependence pattern between drought characteristics in terms of correlation

Drought Index Dependence Measure Duration-Severity P-value for correlation
MSDI Pearson’s Correlation (r) 0.97 1.18e−30
Kendall’s τ 0.89 5.597e−17
Spearman’s rank correlation 0.97 9.28e−30
(ρ)
3.3 Marginal Distribution Selection for Drought

Characteristics
The selection of appropriate marginal distribution is required for drought risk estima-
tion. The study tested several parametric distributions to select the optimum distri-
butions for fitting both duration and severity. The study applied the maximum likeli-
hood (ML) approach for parameter estimation of the distributions. The performance
of each distribution was evaluated using the Kolmogorov–Smirnov (KS) test and
associated p-value at 5% significance level, along with Akaike information criteria
(AIC), and the results are illustrated in Table 5. From Table 5, it can be seen that for
the duration only for the Weibull distribution, the p-value is greater than the signifi-
cance level (0.05). In the case of severity, all the selected distributions have a p-value
greater than the significance level. The plot of CDF and PDF for distributions fitted to
both the drought characteristics are presented in Fig. 4. From the results, the Weibull
distribution was selected for modelling both duration and severity.
Table 5 Performance comparison between fitted probability models

Drought Index Drought Distributions Shape Scale KS p-value (α = AIC
variables 0.05)
MSDI Duration Gamma 1.74 2.21 0.19 0.047 222.24
Weibull 1.35 4.22 0.18 0.083 223.55
Lognormal 1.04 0.82 0.21 0.027 220.40
Exponential 3.85 - 0.23 0.011 227.52
Severity Gamma 1.23 4.50 0.12 0.501 263.33
Weibull 1.09 5.74 0.12 0.511 263.94
Lognormal 1.25 1.00 0.13 0.346 259.93
Exponential 5.54 - 0.14 0.276 262.27
Fig. 4 PDF and CDF plots for the fitted probability distribution of duration and severity
3.4 Analysis of the Joint Distribution of Drought Variables

by Applying Copulas
The joint distribution of the two drought characteristics is formulated using a suit-
able copula function. Five copula functions were evaluated to select the best copula
for bivariate modelling. The maximum pseudo-log-likelihood approach has been
adopted in this study for the estimation of copula parameters. The fitted parameters
and related log-likelihood estimates are listed in Table 6. The Frank copula has the
highest log-likelihood value, followed by the Gumbel-Hougaard copula. The AIC
value of the Frank copula is the least among all the copulas considered in the study.
Further, the results of distance-based statistics AD and IAD are also listed in Table
6.
It is observed that the AD and IAD values are least for Frank copula as well, and
hence the Frank copula is chosen for estimating bivariate drought risk.
186 Rajarshi et al.
Table 6 The estimated parameters of copula and comparison of their performance

Copula Family Copula Parameter LL AIC Sn AD IAD
Clayton 8 = 4.115 33.921 −65.843 0.236 0.397 1.307
Frank 8 = 24.469 61.760 −121.520 0.101 0.391 0.676
Gumbel-Hougaard 8 = 5.873 60.387 −118.774 0.114 0.401 0.746
Gaussian ρ = 0.950 54.464 −106.929 0.128 0.397 0.804
t-copula ϑ = 5.025, 56.266 −108.532 0.115 0.395 0.726
r = 0.962
3.5 Bivariate Frequency Analysis of Droughts
3.5.1 Comparison of the Univariate and Bivariate Drought Return

Period
The joint distributions derived using the selected copula function are further applied
for computing two primary return periods, the ‘OR’ case and the ‘AND’ case.
The average inter-arrival time of drought is 9.75 months for the period of 1981
to 2019. Table 7 presents drought severity and duration for 5, 10, 20, 50, and
100 years of univariate return periods. For instance, in the case of a univariate return
period of 20 years, the severity is greater than 16.68, and the duration is greater
than 10.04 months, respectively. The comparison of univariate with bivariate return
periods computed using the Frank copula is also presented in Table 7.
On the contrary, if we consider joint behaviour of the drought characteristics, it is
observed from Table 7 that the return period for the ‘AND’ case where D ≥ 10.04 and
S ≥ 16.68 is 39.33 years; whereas the return period for ‘OR’ case where D ≥ 10.04
or S ≥ 16.68 is equal to 13.41 years, as presented in Table 7. In a nutshell, in the
case of a 3-month time scale of MSDI, compared to a single variant return period of
20 years, the bivariate return period for the ‘AND’ case is increased by 96.65%, and
if we consider the ‘OR’ case, then the return period decreased by 32.95%. Moreover,
it is noticed that compared to univariate return periods, the bivariate return period for
the ‘AND’ condition is higher, ‘and’ the ‘OR’ condition is lower. It can be inferred
from the analysis that the ‘OR’ condition overestimates the drought risks, and the
Table 7 Univariate and bivariate return periods and their associated return periods in the
Marathwada region
Return Period (T) (years) Duration (months) Severity AN D
TDS OR
TDS
5 6.59 9.92 6.04 4.27
10 8.37 13.34 14.56 7.62
20 10.04 16.68 39.33 13.41
50 12.10 21.01 174.57 29.18
100 13.58 24.22 601.6 54.53
‘AND’ condition underestimates the drought risks. Hence, both these estimates are
useful to understand and manage drought risks in the region.
4 Conclusions
To manage water resources under drought conditions, it is necessary to characterize

droughts using appropriate drought indices and analyze the associated drought risks.
Rather than considering only one aspect, the effect of several drought-causing vari-
ables on the phenomena will give us a broad idea about the spatiotemporal assessment
of drought. Therefore, this study used precipitation and soil moisture-based index for
adequate drought characterization and risk estimation. The drought was modelled
using MSDI at a 3-month time scale for the Marathwada region. The Archimedean
and meta-elliptical families were evaluated for modelling joint distribution of dura-
tion and severity. Univariate and bivariate joint return periods for ‘OR’ and ‘AND’
cases were computed to understand the drought risks in the region. The main findings
from the study are:
i. The Frank copula was ranked best-fit function for constructing parametric joint
distribution for computing MSDI.
ii. MSDI showed good efficacy in monitoring droughts in the region in terms of
identifying the drought events and evaluating their characteristics for the period
of analysis; their values are in concordance with SPI and SSI.
iii. The results of bivariate dependency analysis indicated Frank copula as the
most suitable copula for constructing joint distribution between duration and
severity.
iv. The univariate return period for the study region was higher than the ‘OR’
case joint return period but lower than the ‘AND’ case return period. The study
deduced that the ‘OR’ case joint return period would overestimate the drought
risk while the risk will be underestimated when the ‘AND’ case is considered.
Acknowledgements The authors would like to thank the various organizations that provided
support for the study. Department of Science and Technology (SPLICE–Climate Change
Programme), Government of India, Project #DST/488/CCP/CoE/140/2018 for funding support;
the India Meteorological Department (IMD), the Modern-Era Retrospective analysis for Research
and Applications, Version 2 (MERRA-2) of NASA for providing the datasets used in this study.
References
Agutu, N. O., J. L. Awange, C. Ndehedehe, and M. Mwaniki. 2020. “Consistency of Agricultural

Drought Characterization over Upper Greater Horn of Africa (1982–2013): Topographical, Gauge
Density, and Model Forcing Influence.” Science of the Total Environment 709: 135149. https://
doi.org/10.1016/j.scitotenv.2019.135149.
188 Rajarshi et al.
Ahmadalipour, Ali, Hamid Moradkhani, and Mehmet C. Demirel. 2017. A Comparative Assessment
of Projected Meteorological and Hydrological Droughts: Elucidating the Role of Temperature.
Journal of Hydrology 553: 785–797. https://doi.org/10.1016/j.jhydrol.2017.08.047.
Ane, Thierry, and Cécile. Kharoubi. 2003. Dependence Structure and Risk Measure. The Journal
of Business 76 (3): 411–438. https://doi.org/10.1086/375253.
Chen, Nengcheng, Ronghui Li, Xiang Zhang, Chao Yang, Xiaoping Wang, Linglin Zeng, Shengjun
Tang, Wei Wang, Deren Li, and Dev Niyogi. 2020. “Drought Propagation in Northern China
Plain: A Comparative Analysis of GLDAS and MERRA-2 Datasets.” Journal of Hydrology 588:
125026. https://doi.org/10.1016/j.jhydrol.2020.125026.
Farahmand, Alireza, and Amir AghaKouchak. 2015. A Generalized Framework for Deriving
Nonparametric Standardized Drought Indicators. Advances in Water Resources 76: 140–145.
https://doi.org/10.1016/j.advwatres.2014.11.012.
Ganguli, Poulomi, and M. Janga Reddy. 2012. “Risk Assessment of Droughts in Gujarat Using
Bivariate Copulas.” Water Resources Management 26 (11): 3301–3327. https://doi.org/10.1007%
2Fs11269-012-0073-6.
Gelaro, Ronald, Will McCarty, Max J. Suárez, Ricardo Todling, Andrea Molod, Lawrence Takacs,
Cynthia A. Randles, Anton Darmenov, Michael G. Bosilovich, and Rolf Reichle. 2017. The
Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2).
Journal of Climate 30 (14): 5419–5454. https://doi.org/10.1175/jcli-d-16-0758.1.
Genest, Christian, and Anne-Catherine. Favre. 2007. Everything You Always Wanted to Know about
Copula Modeling but Were Afraid to Ask. Journal of Hydrologic Engineering 12 (4): 347–368.
https://doi.org/10.5040/9781472547675.ch-002.
Hao, Zengchao, and Amir AghaKouchak. 2013. Multivariate Standardized Drought Index: A Para-
metric Multi-Index Model. Advances in Water Resources 57: 12–18. https://doi.org/10.1016/j.
advwatres.2013.03.009.
Hao, Zengchao, and Vijay P. Singh. 2015. Drought Characterization from a Multivariate Perspective:
A Review. Journal of Hydrology 527: 668–678. https://doi.org/10.1016/j.jhydrol.2015.05.031.
Janga Reddy, M., and Poulomi Ganguli. 2012. “Application of Copulas for Derivation of Drought
Severity–Duration–Frequency Curves.” Hydrological Processes 26 (11): 1672–1685. https://doi.
org/10.1002%2Fhyp.8287.
Joe, Harry. 1997. Multivariate Models and Multivariate Dependence Concepts. CRC Press. https://
doi.org/10.1201/9780367803896.
Kim, Tae-Woong, and Juan B Valdés. 2003. “Nonlinear Model for Drought Forecasting Based on a
Conjunction of Wavelet Transforms and Neural Networks.” Journal of Hydrologic Engineering
8 (6): 319–328. https://doi.org/10.1061//ASCE/1084-0699/2003/8:6/319.
Liu, Suning, Haiyun Shi, and Bellie Sivakumar. 2020. “Socioeconomic Drought under Growing
Population and Changing Climate: A New Index Considering the Resilience of a Regional Water
Resources System.” Journal of Geophysical Research: Atmospheres 125 (15): e2020JD033005.
https://doi.org/10.1029%2F2020jd033005.
Liu, Xianfeng, Xiufang Zhu, Yaozhong Pan, Shuangshuang Li, Yanxu Liu, and Yuqi Ma. 2016.
Agricultural Drought Monitoring: Progress, Challenges, and Prospects. Journal of Geographical
Sciences 26 (6): 750–767. https://doi.org/10.1007/s11442-016-1297-9.
McKee, Thomas B., Nolan J. Doesken, and John Kleist. 1993. “The Relationship of Drought
Frequency and Duration to Time Scales.” 17: 179–183. In Proceedings of the 8th Conference on
Applied Climatology.
Mirabbasi, Rasoul, Ahmad Fakheri-Fard, and Yagob Dinpashoh. 2012. “Bivariate Drought
Frequency Analysis Using the Copula Method.” Theoretical and Applied Climatology 108 (1):
191–206. https://doi.org/10.1007%2Fs00704-011-0524-7.
Mishra, Ashok K., and Vijay P. Singh. 2010. A Review of Drought Concepts. Journal of Hydrology
391 (1–2): 202–216. https://doi.org/10.1016/j.jhydrol.2010.07.012.
Nelsen, Roger B. 2007. An Introduction to Copulas. New York: Springer-Verlag. https://doi.org/10.
1007/0-387-28678-0.
Pai, DS, Latha Sridhar, M Rajeevan, OP Sreejith, NS Satbhai, and B Mukhopadhyay. 2014. “Devel-
opment of a New High Spatial Resolution (0.25× 0.25) Long Period (1901–2010) Daily Gridded
Rainfall Data Set over India and Its Comparison with Existing Data Sets over the Region.”
Mausam 65 (1): 1–18. https://doi.org/10.1007/s00382-014-2307-1.
Palmer, Wayne C. 1965. Meteorological Drought. Vol. 30. US Department of Commerce, Weather
Bureau.
Reichle, Rolf H, Clara S Draper, Qing Liu, Manuela Girotto, Sarith PP Mahanama, Randal D
Koster, and Gabrielle JM De Lannoy. 2017. “Assessment of MERRA-2 Land Surface Hydrology
Estimates.” Journal of Climate 30 (8): 2937–60. https://doi.org/10.1175/jcli-d-17-0121.1.
Sathyanadh, Anusha, Anandakumar Karipot, Manish Ranalkar, and Thara Prabhakaran. 2016. Eval-
uation of Soil Moisture Data Products over Indian Region and Analysis of Spatio-Temporal Char-
acteristics with Respect to Monsoon Rainfall. Journal of Hydrology 542: 47–62. https://doi.org/
10.1016/j.jhydrol.2016.08.040.
Shiau, J.T. 2003. Return Period of Bivariate Distributed Extreme Hydrological Events. Stochastic
Environmental Research and Risk Assessment 17 (1): 42–57. https://doi.org/10.1007/s00477-
003-0125-9.
Sklar, M. 1959. Fonctions de Repartition an Dimensions et Leurs Marges. Publications Institute
Statistics University Paris 8: 229–231.
Soľáková, Tatiana, Carlo De Michele, and Renata Vezzoli. 2014. Comparison between Parametric
and Nonparametric Approaches for the Calculation of Two Drought Indices: SPI and SSI. Journal
of Hydrologic Engineering 19 (9): 04014010. https://doi.org/10.1061/(asce)he.1943-5584.000
0942.
Song, Songbai, and Vijay P. Singh. 2010. Frequency Analysis of Droughts Using the Plackett Copula
and Parameter Estimation by Genetic Algorithm. Stochastic Environmental Research and Risk
Assessment 24 (5): 783–805. https://doi.org/10.1007/s00477-010-0364-5.
Willems, Patrick. 2000. Compound Intensity/Duration/Frequency-Relationships of Extreme Precip-
itation for Two Seasons and Two Storm Types. Journal of Hydrology 233 (1–4): 189–205. https://
doi.org/10.1016/s0022-1694(00)00233-x.
Yevjevich, Vujica M. 1967. “Objective Approach to Definitions and Investigations of Continental
Hydrologic Droughts.” Colorado State University. Libraries. https://doi.org/10.1016/0022-169
4(69)90110-3.
Zhang, Lan, and V. P. Singh. 2019. Copulas and Their Applications in Water Resources Engineering.
Cambridge University Press, Cambridge. https://doi.org/10.1017/9781108565103.
Zhang, Xiang, Renee Obringer, Chehan Wei, Nengcheng Chen, and Dev Niyogi. 2017. “Droughts
in India from 1981 to 2013 and Implications to Wheat Production.” Scientific Reports 7 (1): 1–12.
https://doi.org/10.1038/srep44552.
High-Quality Historical Flood Data
Reconstruction in Bangladesh Using
Hidden Markov Models
Max Mauerman, Elizabeth Tellman, Upmanu Lall, Marco Tedesco,

Paolo Colosio, Mitchell Thomas, Daniel Osgood, and Arifuzzaman Bhuyan
Abstract Historical data on flood hazard is critical for the design of climate risk
management policies. Parametric weather insurance typically requires at least 20–
30 years of historical data for accurate risk pricing. However, the current sources
of satellite data on flood inundation in Bangladesh are limited. Although passive
microwave (PMW) measurements reach back to 1992, they are imprecise and at
coarse spatial resolution (25–3.125km2 ). Sentinel-1 measurements are more precise
(at 10m2 resolution) but are only consistently available from 2017 onwards. We
present a method for reconstructing the high-quality signal of Sentinel-1 data for
pre-2017 years using its statistical relationship to PMW—a “data fusion” process.
Our data fusion is based on a Bayesian Hidden Markov Model (HMM), in which both
Sentinel-1 and PMW are modelled as realizations of an unobserved, time-dependent
discrete variable representing flood state. Evaluated against the actual Sentinel-1 in
M. Mauerman (B) · D. Osgood

International Research Institute for Climate and Society, Columbia University, New York, NY,
USA
E. Tellman
School of Geography, Development and Environment, University of Arizona, Tucson, Arizona,
USA
U. Lall
Department of Earth and Environmental Engineering, Columbia University, New York, NY, USA
M. Tedesco
Lamont-Doherty Earth Observatory, Columbia University, New York, NY, USA
P. Colosio
Department of Civil, Environmental, Architectural Engineering and Mathematics, University of
Brescia, Brescia, Italy
M. Thomas
Department of Mathematics, Columbia University, New York, NY, USA
A. Bhuyan
Bangladesh Flood Forecasting and Warning Centre, Dhaka, Bangladesh
192 M. Mauerman et al.
Sylhet division, the “fused” data series has a Spearman correlation of 65%. The fused
series also has a 71% Spearman correlation with stream gauge measurements of water
levels. On both metrics of correlation, the simulated series achieves an improvement
over using PMW alone. We discuss how this simulated data could be used to improve
parametric flood insurance.
Keywords Flooding · Insurance · Bangladesh · Data fusion · Remote sensing
1 Introduction
1.1 Overview
Floods have a greater human cost than any other disaster, having affected 2.3 billion
people globally from 1995 to 2015 (UNISDR 2015). This cost falls disproportion-
ately on marginalized populations (Winsemius et al. 2018; Tellman et al. 2020)—in
the world’s 77 lowest-income countries, 92% of disaster costs are not covered by
international assistance (RMS 2017). There is a need for financial risk management
strategies to cover these hazards—only 3% of this financing gap is covered by any
kind of insurance (RMS 2017). Bangladesh in particular is among the world’s most
flood-prone countries—flood extent covers 25% of the country’s area in a normal
year and can reach over 60% in an extreme year (Mirza 2002).
Disaster insurance has been recognized as a way to mitigate catastrophic setbacks
like floods and help the economically vulnerable escape poverty traps (Barnett et al.
2008). In particular, agricultural index insurance—in which payouts are based on
sensor measurements of areas’ weather hazards rather than individual claims assess-
ments—has been promoted by many international organizations and foreign assis-
tance agencies (Osgood and Shirley 2012; Benami et al. 2021) as a sustainable way
to provide affordable financial protection in developing nations.
However, the potential advantages of index insurance depend on the availability of
timely climate data that accurately reflects populations’ exposure to weather hazards
(Benami et al. 2021). For example, in Bangladesh, flood risk estimation has been
typically based on stream gauge water levels, but the stream gauge network is limited
in its coverage and cannot convey the full picture of floods’ inundated area. A work-
shop with government, NGO, and private sector stakeholders in Bangladesh identi-
fied inadequate risk exposure and monitoring data as the main obstacle to scaling up
weather index insurance, particularly flood insurance. (ICCAD 2013).
Advances in radar-based satellite flood mapping technology could help meet this
need, but the historical record of these new high-resolution products is short, limiting
their utility for insurance. There is a need for statistical methods to reconstruct
high-quality historical flood risk information in countries like Bangladesh. Here, we
present a data fusion approach based on Bayesian Hidden Markov Models, which
combines a high-resolution satellite with a short record (2017–present) and a lower
High-Quality Historical Flood Data Reconstruction in … 193
spatial resolution satellite with a longer record (1992–present) to design flood risk
insurance products.
The first part of our paper describes the flood data sources used, along with their
practical advantages and disadvantages. The second part introduces the principle of
Hidden Markov Models and demonstrates how a well-identified HMM can be fitted
to our Bangladesh flood data, using Bayesian methods. Finally, the third part of our
paper demonstrates how our reconstructed flood data series has a high degree of corre-
lation with other, independently measured proxies of flood risk, and describes how it
could be used to improve index insurance—ultimately leading to risk management
products that are both more accurate and more sustainably priced.
1.2 Disaster Index Insurance
Parametric disaster insurance, also known as index insurance, is a type of insurance

product designed to pay out based on a proxy for weather hazards, rather than on
individual claims assessments as in conventional indemnity-based insurance. The
most common such proxy is satellite data on weather conditions, which has the
advantages of being timely, low- or no-cost, and global in their coverage (Osgood
and Shirley 2012; Benami et al. 2021).
A typical satellite-based index insurance contract is designed to cover an entire
geographically contiguous area, such as a county-level administrative unit. These
areas are known as the “unit areas of insurance” (UAIs). Satellite weather data is
typically spatially averaged over the UAI to arrive at a single value for each insured
area. This value is compared against the contract “trigger”, an agreed-upon threshold
beyond which the insurance product will pay out. Trigger values are often defined
in reference to some historical return period; for instance, many index insurance
products are tuned such that they are expected to pay out one out of every 5 or one
out of every 10 years (Benami et al. 2021; Barnett et al. 2008). If the insurance
contract triggers in a given UAI, all claimants in that area receive a payout.
A common issue with index insurance is “basis risk”; that is, the mismatch between
insurance payouts and claimants’ actual losses. Basis risk is a particular issue for
satellite-based insurance, where sensor biases and imprecisions in measurement
mean products may fail to pay out during critical loss years (Osgood et al. 2018).
There is a growing body of literature on how to quantify basis risk in index insurance
and adjust contracts to minimize it (Osgood and Shirley 2012; Benami et al. 2021;
Carter et al. 2017).
Satellite-based index insurance has most commonly been applied to drought or
excess of rainfall. There is increasing interest in applying similar methods to flood
insurance—Bangladesh is the first country to pilot an index insurance product for
floods (United News of Bangladesh 2020) that is based on direct estimates of inun-
dated area rather than on rainfall (as in CCRIF 2015). However, satellite data on flood
frequency and inundation extent of suitable quality for insurance remains limited.
1.3 Satellite Flood Estimation
Satellites can detect the inundated area of large flood events in rural areas by classi-
fying areas that are distinguishable as water from space (Pekel et al. 2016; Cooley
et al. 2017; Ji et al. 2018; Feng et al. 2019; Jones 2019). Radar satellite observa-
tions are particularly useful for flood mapping because clouds do not obscure their
observations.
However, among the current radar-based remote sensing flood products, there is
a tradeoff between accuracy, precision, and length of record, as detailed in Sect. 2.
Products based on passive microwave estimates of soil moisture have a record of
several decades, but have a very coarse resolution. In contrast, products based on
synthetic aperture radar offer a much higher resolution—however, regular measure-
ments only exist from early 2017 onward. Additionally, passive microwave sensors
may be more prone than synthetic aperture sensors to false positives from non-
flood sources of soil moisture, such as agricultural irrigation. Because rice irrigation
requires standing water, rice fields irrigated during the dry season can often appear to
be inundated both from the ground and consequently from satellites (Thomas et al.
in pub).
1.4 Data Fusion
The tradeoff between temporal length and spatial resolution described above presents
a quandary for index insurance projects, as an ideal index not only has low basis
risk (i.e. its measurements accurately reflect the hazard that claimants face on the
ground) but also has enough historical data to usefully inform risk pricing. Since a
typical index insurance contract is designed to pay out once out of every 5–10 years
on average, having a precise estimate of the magnitude of rare events is critical.
Uncertainty in this regard could lead to unsustainably priced financial instruments
and / or be passed on to the buyer in the form of higher risk premia (Osgood and
Shirley 2012; Bell et al. 2013).
A potential way to combine the advantages of short, high-quality satellite sources,
and longer, noisier satellite sources is to use statistical methods to reconstruct what
the high-quality source would have measured in years before its record, using the
longer but noisier source as a proxy. This class of solutions is known as “data fusion”,
and has many applications throughout climate science (Bellone et al. 2000; Hughes
et al. 1999; Robertson et al. 2003; Xie et al. 2018). Zeng et al. (2020) present a data
fusion method to combine synthetic aperture radar and passive microwave satellites,
but this method is based on fitting a quadratic function to a very small (fewer than
50) number of data points—and thus is likely prone to overfitting. A promising
alternative approach is to treat the parametric relationship between the two series
as a time-varying process, the dynamics of which are guided by a simple rule—so-
called Hidden Markov Models (HMMs). HMMs have been used by climate scientists
to reconstruct phenomena like wind speed and rainfall (e.g. Robertson et al. 2003);
here, we apply them to the problem of reconstructing division-level fractional flooded
area.
2 Data
We consider two sources of satellite data on fractional flooded area for this paper:
Sentinel-1 (“S1”) and passive microwave SSM/I (“PMW”). For the reasons detailed
below, we believe that S1 is less likely to suffer from biases in extreme event estima-
tion than PMW—however, its record is significantly shorter. This is what motivates
the application of data fusion techniques.
2.1 Sentinel-1
The Sentinel-1 satellite product is a two-satellite constellation of Synthetic-Aperture

Radar (SAR) sensors launched by the European Space Agency in April 2014 and
April 2016, respectively. Typical of SAR sensors, Sentinel-1 features the ability
to capture images through clouds, haze, and at night-time (Rahman and Thakur
2018). Sentinel-1 data usable for insurance is available over Bangladesh from 2017
to present at a current average revisit time of ~ 4.2 days, although this comes as a
result of a primarily 2- and 10-day overpass schedule. Spatial resolution is 10 m. Data
in this study is used in two transmit/receive radar polarizations—Vertical–Vertical
(VV) and Vertical-Horizontal (VH)—and two overpass directions—Ascending and
Descending.
We use a flood detection algorithm developed by adapting the Sentinel-1 statistical
deviation flood algorithm proposed in DeVries et al. (2020). An adapted algorithm is
proposed which adds a baseline selection method, an additional threshold on the VH
band, and spatial smoothing. A dry-baseline period is derived from the minimum 45-
days of a soil moisture time series over the studied area. We then compute a mean and
standard deviation image for both polarizations and both overpass directions from
Sentinel-1 images within this baseline period. From these images, a z-score image
is computed at every time step where Sentinel-1 data is available 2017–Present. A
binary (water or not water) water map is created from the z-score image by marking
any pixel as water if the vv or vh z-score value is < –2.5. Finally, spatial smoothing
is applied, at both 2- and 4-pixel radii. A pixel is classified as flooded if either the
2-pixel radius mean is > 0.6 or the 4-pixel radius mean is > 0.4.
Fractional flooded area over an administrative area is computed by dividing the
total sum of water in the region by region area. A time series of fractional flooded area
can accordingly be calculated at any administrative level in Bangladesh by iterating
over every Sentinel-1 overpass. For this study, we use estimates by division for three
divisions in the north: Rajshahi, Sylhet, and Rangpur.
2.2 Passive Microwave SSM/I
Passive microwave (PMW) sensors measure the electromagnetic radiation naturally

emitted by the Earth surface, expressed as brightness temperature (ULABY 1982).
Because of their sensitivity to wet and dry surfaces, different approaches have been
proposed in the past aimed at monitoring large flood events continuously (CHOUD-
HURY 1989; Galantowicz 2002; Groeve 2010; Tangdamrongsub et al. 2021). In fact,
the peculiarities of PMW spaceborne records are their long and continuous temporal
coverage (at least every other day since 1979) and the large spatial coverage (global).
On the other hand, the relatively coarse spatial resolution of 25 km2 has always been a
major limitation of this dataset. Recently, in the framework of the NASA MeASUREs
project, a new gridding algorithm called rSIR technique (Brodzik and Long 2016)
has been applied to PMW records to produce a new dataset at the enhanced resolution
of 3.125 km2 .
Here, we apply a variation of the algorithm proposed by Groeve (2010) to
the enhanced resolution PMW data from the Special Sensor Microwave / Imager
(SSM/I) and the Special Sensor Microwave Imager / Sounder (SSMI/S) at Ka- band
(37 GHz, horizontal polarization) collected during the ascending pass of the satellite
between 1992 and 2019. The algorithm is based on the assumption that spatially
close pixels have similar land surface properties and on the choice of a dynami-
cally computed calibration brightness temperature to detrend the time series from
temperature seasonality.
The measured (M) brightness temperature value of each pixel is considered
as linear combination of water and land brightness temperature contributions,
proportionally to the fractional area, and can be expressed as:
Tb,measur ement = (1 − w)Tb,land + wTb,water

= Tmeasur ement (1 − w)εland,measur ement + wεwater (1)
where w is the water fractional area, εwater the water emissivity, εland the land
emissivity and Tmeasurement the physical temperature at the measured pixel.
The measured brightness temperature (M, “wet” or “dry”) is normalized with
respect to a calibration brightness temperature value (C), chosen to represent the
driest condition of a certain area (i.e. w = 0) as
Tb,calibration = Tb,land = Tcalibration εland,calibration (2)
where Tcalibration is the physical temperature of the calibration pixel.

For nearby pixels, we consider valid the fundamental assumption of equal
emissivities εland,measur ement ≈ εland,calibration ≈ εland and equal temperatures
(Tcalibration = Tmeasur ement ).
Finally, we compute the calibrated signal s (or M/C) as the ratio between the
measurement and the calibration brightness temperatures. This quantity is dependent
on the water fraction only, increasing when w decreases.
s = M/C = Tb,measur ement /Tb,calibration ≈ 1 − w(1 − εwater /εland ) = f (w) (3)
In order to automatically select the calibration pixel, Groeve (2010) suggested

the 95th percentile in a 7 by 7 pixels moving window (30,625 km2 ). Here we use
the 95th percentile in a 14 by 14 pixels moving window, according to preliminary
analysis of the time series of calibration values.
Groeve (2010) defined as flood magnitude the number of standard deviations from
the mean of the signal s. According to this definition, regular (large) flood events are
identified with flood magnitude equal to 2. Here we consider a flood magnitude of 2
as threshold to identify flood events.
As with Sentinel-1, these pixel-level binary indicators of flood presence are
spatially averaged to arrive at an estimate of fractional flooded area for a region
(here, three divisions in the north: Rajshahi, Sylhet, and Rangpur).
2.3 Tradeoffs Between Sentinel-1 and Passive Microwave
Passive microwave sensors have at least three potential sources of noise for flood
detection—soil moisture, rainfall, and coarse spatial resolution—that are less of a
problem in active radar sensors like Sentinel-1.
As described above, passive microwave sensors detect water by changes in bright-
ness temperature. Passive microwave sensors are typically used to assess soil mois-
ture, in part because they are so sensitive to wet soils (Fang et al. 2018; Sabaghy
et al. 2018). Wet soil (that is not flooded or has standing water present) can cause
a similar drop in brightness temperature to inundated area. Thus, it can be difficult
to distinguish the signal from an irrigated and actively cultivated field to one that is
simply waterlogged, from one with standing water.
Similarly, heavy rainfall can interfere with the signature of surface conditions in
PMW. Microwave algorithms over land depend on scattering signals, and the signals
of atmospheric precipitation and surface type can be difficult to distinguish (Scheel
et al. 2011).
Regarding spatial resolution, Fig. 1 illustrates the size of the smallest resolvable
flood event each sensor is capable of detecting. Passive microwave resolution is
3.125km2 ; in contrast, Sentinel-1 resolution is 10m2 . The significantly higher reso-
lution of Sentinel-1 means it is much more able to discriminate the signal of flood
events from non-flood events. Similarly, floods in areas adjacent to permanent bodies
of water may be undetectable by passive microwave, as the signature of flooding is
indistinguishable from that of permanent water for co-located pixels.
To illustrate the advantages of higher spatial resolution in practical terms, Fig. 2
illustrates 6 different satellite measurements of the same flood event in northern
Bangladesh. We can see how in Sentinel-1, riverine overflow is clearly resolvable,
while in passive microwave, the entire study area comprises less than a single pixel.
The biases described above are apparent in a comparison of each sensor’s esti-
mated fractional flooded area in Sylhet division (Fig. 3). All data are aggregated to
Fig. 1 Comparison of satellite flood products’ resolution and measurement frequency. Figure
reproduced from Tellman et al. (2021)
Fig. 2 Comparison of satellites’ view of a flood event in northern Bangladesh
the dekadal (10-day) time scale by taking the maximum value over each dekad—a
convention followed for the rest of this paper.
In the planting season, PMW exhibits erroneous spikes from seasonal irrigation
around December to January, whereas Sentinel-1 does not. Likewise, the tendency of
the PMW signal to saturate leads to a much more “jagged” time series than Sentinel-1
during the rainy season, with transient peaks and valleys in fractional flooded area.
Fig. 3 Comparison of S1 and PMW SSM/I flooded area in Sylhet
Since index insurance is based on the accurate measurement of the magnitude of

extreme events, these biases present a major practical problem.
However, visualizing the joint distribution of the two series over each quarter of
the year (Fig. 4) suggests a way forward. We can see that within each quarter, the
linear relationship between the two series is fairly strong, even as the magnitude
of the relationship varies over seasons. This suggests that data fusion models which
allow the relationship between the predictor and predicted series to vary as a function
of time could be promising. The following section explores such a model in detail.
Fig. 4 Joint distribution of S1 and PMW flooded area by quarter of the year
2.4 Validation Data
To establish a benchmark for the reliability of our data fusion model (and its improve-
ment over existing satellite products), we need an external source of data on the timing
and severity of flood events to compare it to. While no data source is “perfect”, events
that are independently reflected in multiple sources are more likely to reflect actual
weather risk phenomena and not idiosyncratic noise or bias.
For this paper, our validation data comes from a dataset of stream gauge water
levels obtained from the Bangladesh Flood Forecasting and Warning Centre (FFWC).
This data is available from 2008 to 2019 and is measured on a biweekly frequency
(with some irregularity) over the rainy season (May–October) only. Of the 53 gauges
available, we select 1 in each division for comparison on the basis of the highest
maximum value over the observation period, under the intuition that this gauge is
likely the most representative of flood events in the district. As with the satellite
data, the stream gauge data is temporally aggregated up to the dekadal (10-day)
level, taking the maximum value over each dekad.
3 Methodology
3.1 Model Overview
A Markov process is a simple rule for describing a phenomenon which depends on

its past values. In its simplest form, a first-order Markov process can take a finite
number of discrete states, and its state in period t depends only on its state in t-1.
A “transition probability matrix” describes the likelihood of going from state i in
period t-1 to state j in period t, as in Fig. 5.
Weather and climate dynamics have considered Markov processes for rainfall
(wet/dry states), temperature (continuous variables with autoregressions), and other
variables, with transition probabilities (or autoregressions) that may be seasonally
variable (Robertson et al. 2003; Bellone et al. 2000; Xie et al. 2018; Hughes et al.
1999). In the machine learning literature, Hidden Markov Models (HMMs) were
introduced where we do not observe the Markov state directly; rather, the dynamics
of the state variable (continuous or discrete) are believed to be influenced by a hidden
or latent variable. The underlying dynamics, including the transition probabilities
of the discrete hidden states need to be inferred directly from the observed state
variable alone. For our application, we consider that the hidden state may refer to a
flood signature over the domain that is seen by both the Sentinel-1 and the PMW.
This hidden state or signature influences the relationship or correspondence between
these two variables.
In our study’s observation model, the distribution of Sentinel-1 in each state is
described by a Gaussian distribution. In a univariate model, the distribution of S1 over
each state would have fixed mean and variance. However, this is a data fusion model,
Fig. 5 Example of transition

probability matrix
with two variables—S1 (the observations) and PMW (the predictor). Therefore, the
mean of the S1 distribution in each state is not fixed, but is instead a scalar function of
PMW at each time step. This is functionally equivalent to running a linear regression
of S1 on PMW where the slope coefficient is allowed to vary by Markov state.
Formally, the observation model is:
S1t = N (βs P M Wt , σs ) (4)
Where S1t is the estimated Sentinel-1 fractional flooded area at time t, PWT t is
the passive microwave fractional flooded area at time t, and βs and σs are unique
constants for each hidden state s.
We chose a Gaussian observation model because it is straightforward to model
one series as a linear function of another in this framework. In practice, however,
the joint distribution of S1 and PMW is often non-normal (as with many climate
phenomena, there is a long tail of rare extreme events), so it would be valuable to
explore other link functions in future work. (Running the model in logarithms was
not found to improve its fit.)
We consider a model with three hidden states for this paper. A 3 × 3 transition
matrix describes the probability of transitioning from state i in period t-1 to state j in
period t (as detailed above). The model also requires an estimate of the initial state
probability at t = 0.
We fit the model using a Hamiltonian Monte Carlo Bayesian sampling method in
the Stan statistical package (Gelman et al. 2015); full details of the Bayesian model
priors are available in the appendix. This method gives us a full estimated distribution
for the values of each parameter in the model: the transition probabilities, the initial
state probability, and the mean and variance of the observation model at each hidden
state. We keep the priors as weak as possible; the only restriction given is that the
observation distribution must be bounded by [0,1] so that the predicted fractional
flooded area has a sensical interpretation.
3.2 Data Fusion Method
Once we have fit the model, we can use it for retrospective prediction. The first step
of prediction is to fit the likeliest hidden state path to the PMW data we have. To do
this, we use the Viterbi algorithm (Forney 1973), which computes the most likely
sequence of hidden states Z* iteratively, from the initial probability at t = 0 to the
final period T:

∗
Z = arg max p(S11:T ) . . . , arg max p(Z T |S11:T ) (5)
z1 zT
We have to make one additional assumption, which is that the Viterbi path learned
from PMW alone (since that is all we observe pre-2017) is a reasonable proxy for
the Viterbi path in the joint model. As long as we think both sensors are picking up
the same underlying weather dynamics, this is a reasonable assumption.
The second step of prediction is to simulate what the Sentinel measurement would
have been at each time step, given the likeliest hidden state and the PMW observation.
Since there is uncertainty in both of these steps—both in the posterior distribution
of estimated parameters as well as the underlying observation distribution—it is
informative to run an ensemble of predictions. Figure 6 summarizes the entire data
fusion process.
Fig. 6 Data fusion method schematic

4 Results
We fit our model separately to three administrative divisions in the north of

Bangladesh—Rajshahi, Sylhet, and Rangpur. For each, we fit the model on S1 and
PMW data from 2017 to 2019, then use the model to estimate fractional flooded area
back to 2008, the beginning of the FFWC stream gauge record. The results shown
in this section focus on Sylhet; the conclusion discusses the model’s applicability to
other areas.
4.1 Model Identification
The first question we can ask is whether the model parameters appear to be well-
identified. In a Monte Carlo-based programme like Stan, identification is typically
estimated through convergence statistics like R-hat (Gelman et al. 2015). Stan solves
for the posterior distribution of model parameters by running a number of parallel
simulations (“chains”) starting from different points in the solution space. We would
expect the chains to converge eventually if the model is well-identified. We do in fact
see that in our model (Fig. 7).
Figure 8 shows the median Viterbi path (likeliest path of hidden states) as
calculated from the model’s posterior distribution.
Figure 8 depicts what the final simulated S1 series looks like, 2008–2019. The
median of the prediction ensemble is shown in dark grey, and the 25th–75th quantile
spread is shown in light grey. The actual S1 observation record is shown in green
(Fig. 9).
For simplicity’s sake, we will collapse the prediction distribution at each time
step to its median for the subsequent evaluation steps. We consider two metrics for
Fig. 7 Convergence graphs for HMM transition matrix parameters

Fig. 8 Viterbi path for Sylhet superimposed on flood data
Fig. 9 Estimated Sentinel-1 probability distribution from data fusion model, Sylhet
model evaluation: Its quality of fit to the Sentinel-1 data (predictive accuracy) and
its correlation with independently measured data on flood risk from stream gauges
(external validity). On each of these metrics, we can compare our data fusion series
against PMW, to establish that it improves on existing instruments.
Table 1 Correlation between

n = 108 Data fusion model SSM/I alone
data fusion results and
(Standard error) (Standard error)
observed Sentinel-1
Pearson correlation 0.72 0.55
with S1 (0.10) (0.15)
Spearman correlation 0.64 0.45
with S1 (0.10) (0.15)
4.2 Predictive Accuracy
The first evaluation step is to establish is how well-simulated S1 correlates with

actual S1, shown in Table 1.
4.3 External Validity
The second evaluation step is to establish external validity; in other words, how well
our simulated series correlates with independently measured sources of data on flood
risk. Our external risk data comes from the FFWC dataset on directly measured water
level from stream gauges, 2008–2019, described in Sect. 2. Table 2 describes these
results.
Overall, we can see that our data fusion model is both an accurate (Spearman
correlation of 64%) predictor of S1 and corresponds strongly (Spearman correlation
of 71%) with stream gauge data in Sylhet—and that on both counts, it represents an
improvement over using passive microwave alone, although this improvement is not
statistically significant.
Our data fusion model did not perform as well in the other two divisions studied,
Rajshahi, and Rangpur. The model is not as well-identified and the resulting predic-
tions do not perform as well on either of our two evaluation metrics. This may be
due to the time scale of the variation in the linear relationship between S1 and PMW
in these areas. In Sylhet, erroneous spikes in the PMW series occur on an irregular
basis and are relatively transient, lasting only 2 or 3 dekads. An HMM model with a
dekadal time scale is well suited to identifying and correcting for these anomalies.
In contrast, the other two divisions exhibit a relationship between S1 and PMW that
is relatively constant over the season. In these cases, the HMM model struggles to
Table 2 Correlation between

n = 288 Data fusion model SSM/I alone
data fusion results and stream
(Standard error) (Standard error)
gauge water level
Pearson Corr. with 0.51 0.47
Stream Gauge (0.15) (0.15)
Spearman Corr 0.71 0.63
with Stream Gauge (0.08) (0.10)
fit the data, classifying most observations in a single state. Models that allow the
relationship between the two series to vary on a quarterly or seasonal relationship
may be more suited to the conditions in these areas.
5 Policy Application
The ultimate goal of this data fusion effort is to generate better data for index insurance
design. One aspect of a quality index insurance data source is low basis risk—that
is, that it accurately reflects the actual harvest risk from weather that claimants expe-
rience. We have already demonstrated that both the actual and simulated Sentinel-1
fractional flooded area have a stronger correlation with stream gauge data than passive
microwave data alone does—and that this better correlation likely relates to the phys-
ical properties of the sensor, as described in Sect. 2 (for further discussion of this
point, see Thomas et al. (in pub)).
However, another important aspect of a quality index insurance data source—and
the primary motivation for this paper—is its length. Most index insurance contracts
are calibrated to pay out during extreme events—for instance, 1-out-of-5-year or 1-
out-of-10-year floods. Some projects estimate the magnitude of such events directly,
using empirical quantiles of the data record; others use parametric methods like the
extreme value distribution (Chowdhury et al. 1991). Either way, Sentinel-1’s four
years of data are insufficient for this purpose. Data fusion unlocks a much longer
record, going back to 1992.
The following graph illustrates the information that data fusion adds in Rajshahi
division—we can see that some of the biggest flood events in the modern record,
such as 2007, lie outside Sentinel-1’s record. Without such years in the record, any
potential index insurance product would likely be miscalibrated against the true level
of risk. Figure 10 illustrates the maximum flood extent in Sylhet in the observed
Sentinel-1 data (green series) and in the fused data (grey series).
Ultimately, we expect the information from our data fusion method should make
flood index insurance products both more accurately and more affordably priced.
6 Conclusion
In sum, we have demonstrated that a Hidden Markov Model applied to fractional

flooded area data from the Sentinel-1 and SSM/I sensors is a promising approach
for reconstructing high-quality historical flood risk information in parts of the world
where there are major data gaps, like Bangladesh. The model is well-identified, yields
demonstrable improvements over existing products, and has clear utility for policy
applications like insurance.
Other data fusion models may be able to improve over HMM in this setting,
and approaches that model variation in the functional relationship over a seasonal
Fig. 10 Comparison of highest fractional flooded area in the Sentinel-1 observation record vs. the
data fusion results
timescale—rather than 10-day periods—may be a particularly valuable direction for

future work. Additionally, this paper takes the administrative division as the unit of
analysis for simplicity’s sake, but spatial aggregation may be better related to areas
of similar climatology and hydrology—another direction for future work. Finally, as
mentioned in the previous section, more work could be done to quantify the financial
benefits of the reduction in uncertainty and basis risk that this model provides in an
insurance setting.
As climate change portends an increase in the likelihood and severity of major
floods in vulnerable countries like Bangladesh, a better understanding of the historical
record is key for policymakers. Our work in this paper not only illustrates the past,
but also provides the material to design better financial instruments against risk in
future.
Acknowledgements Stream gauge data is provided courtesy of the Bangladesh Flood Forecasting
and Warning Centre (FFWC).
This work is undertaken as part of the Columbia World Project, ACToday, Columbia University
in the City of New York. ACToday funding was used to support IRI staff time and conference fees.
Model fitting was done with the Stan statistical package (https://mc-stan.org/) and Stan code
adapted from Luis Damiano, Michael Weylandt and Brian Peterson’s Hidden Markov Model
routines (https://luisdamiano.github.io/BayesHMM).
Appendix
Bayesian Model Priors
Observation model priors:
βs = U ni f or m(0, 1)
σs = U ni f or m(0, 1)
Transition model priors:
Pi, j = Dirichlet(α1 = 1, α2 = 1, α3 = 1)
Initial distribution priors:
P0 = Dirichlet(α1 = 1, α2 = 1, α3 = 1)
References
Barnett, Barry, Christopher Barrett, and Jerry R. Skees. 2008. Poverty Traps and Index-Based Risk
Transfer Products. World Development 36 (10): 1766–1785.
Bell, Andrew R., Daniel E. Osgood, Benjamin I. Cook, Kevin J. Anchukaitis, Geoffrey R. McCarney,
Arthur M. Greene, Brendan M. Buckley, and Edward R. Cook. 2013. Paleoclimate Histories
Improve Access and Sustainability in Index Insurance Programs. Global Environmental Change
23 (4): 774–781. https://doi.org/10.1016/j.gloenvcha.2013.03.003.
Bellone, Enrica, James P. Hughes, and Peter Guttorp. 2000. A Hidden Markov Model for Down-
scaling Synoptic Atmospheric Patterns to Precipitation Amounts. Climate Research 15 (1): 1–12.
https://doi.org/10.3354/cr015001.
Benami, Elinor, Zhenong Jin, Michael R. Carter, Aniruddha Ghosh, Robert J. Hijmans, Andrew
Hobbs, Benson Kenduiywo, and David B. Lobell. 2021. Uniting Remote Sensing, Crop Modelling
and Economics for Agricultural Risk Management. Nature Reviews Earth & Environment 2 (2):
140–159. https://doi.org/10.1038/s43017-020-00122-y.
Brodzik, M. J., and D. G. Long. 2016. “MEaSUREs Calibrated Enhanced-Resolution Passive
Microwave Daily EASE-Grid 2.0 Brightness Temperature ESDR, Version 1.” NASA National
Snow and Ice Data Center DAAC. https://doi.org/10.5067/MEASURES/CRYOSPHERE/
NSIDC-0630.001.
Carter, Michael, Alain de Janvry, Elisabeth Sadoulet, and Alexandros Sarris. 2017. Index Insurance
for Developing Country Agriculture: A Reassessment. Annual Review of Resource Economics 9
(1): 421–438. https://doi.org/10.1146/annurev-resource-100516-053352.
CCRIF. 2015. “The CCRIF Excess Rainfall (XSR) Model.” https://www.ccrif.org/sites/default/
files/publications/technical-materials/CCRIF_Excess_Rainfall_Model_2016_web.pdf.
Choudhury, Bhaskar J. 1989. “Monitoring Global Land Surface Using Nimbus-7 37 GHz Data
Theory and Examples.” International Journal of Remote Sensing 10 (10): 1579–1605. https://
doi.org/10.1080/01431168908903993.
Chowdhury, Jahir, Jery Stedinger, and Li-Hsiung Lu. 1991. “Goodness-of-Fit Tests for Regional
Generalized Extreme Value Flood Distributions.” Water Resources Research—WATER RESOUR
RES 27 (July): 1765–76. https://doi.org/10.1029/91WR00077.
Cooley, Sarah W., Laurence C. Smith, Leon Stepan, and Joseph Mascaro. 2017. Tracking Dynamic
Northern Surface Water Changes with High-Frequency Planet CubeSat Imagery. Remote Sensing
9 (12): 1306. https://doi.org/10.3390/rs9121306.
DeVries, Ben, Chengquan Huang, John Armston, Wenli Huang, John W. Jones, and Megan W.
Lang. 2020. “Rapid and Robust Monitoring of Flood Events Using Sentinel-1 and Landsat Data
on the Google Earth Engine.” Remote Sensing of Environment 240 (April): 111664. https://doi.
org/10.1016/j.rse.2020.111664.
Fang, Bin, Venkataraman Lakshmi, Rajat Bindlish, and Thomas J. Jackson. 2018. “Downscaling
of SMAP Soil Moisture Using Land Surface Temperature and Vegetation Data.” Vadose Zone
Journal 17 (1): 170198. https://doi.org/10.2136/vzj2017.11.0198.
Feng, Dongmei, Colin J. Gleason, Xiao Yang, and Tamlin M. Pavelsky. 2019. Comparing Discharge
Estimates Made via the BAM Algorithm in High-Order Arctic Rivers Derived Solely From Optical
CubeSat, Landsat, and Sentinel-2 Data. Water Resources Research 55 (9): 7753–7771. https://
doi.org/10.1029/2019WR025599.
Forney, G.D. 1973. The Viterbi Algorithm. Proceedings of the IEEE 61 (3): 268–278. https://doi.
org/10.1109/PROC.1973.9030.
Galantowicz, J. 2002. High-Resolution Flood Mapping from Low-Resolution Passive Microwave
Data. IEEE International Geoscience and Remote Sensing Symposium. https://doi.org/10.1109/
IGARSS.2002.1026161.
Gelman, Andrew, Daniel Lee, and Jiqiang Guo. 2015. Stan: A Probabilistic Programming Language
for Bayesian Inference and Optimization. Journal of Educational and Behavioral Statistics 40
(5): 530–543. https://doi.org/10.3102/1076998615606113.
De Groeve, Tom. 2010. Flood Monitoring and Mapping Using Passive Microwave Remote Sensing
in Namibia. Geomatics, Natural Hazards and Risk 1 (1): 19–35. https://doi.org/10.1080/194757
01003648085.
Hughes, J.P., P. Guttorp, and S.P. Charles. 1999. A Non-Homogeneous Hidden Markov Model for
Precipitation Occurrence. Journal of the Royal Statistical Society: Series C (applied Statistics)
48 (1): 15–30. https://doi.org/10.1111/1467-9876.00136.
ICCAD. 2013. “Weather Index Insurance: Lessons Learned and Best Practices for
Bangladesh.” http://icccad.net/wp-content/uploads/2015/12/Weather-Index-Insurance-lessons-
and-best-practices.pdf.
Ji, Luyan, Peng Gong, Jie Wang, Jiancheng Shi, and Zhiliang Zhu. 2018. Construction of the
500-m Resolution Daily Global Surface Water Change Database (2001–2016). Water Resources
Research 54 (12): 10270–10292. https://doi.org/10.1029/2018WR023060.
Jones, John W. 2019. Improved Automated Detection of Subpixel-Scale Inundation—Revised
Dynamic Surface Water Extent (DSWE) Partial Surface Water Tests. Remote Sensing 11 (4):
374. https://doi.org/10.3390/rs11040374.
Mirza, Monirul. 2002. Global Warming and Changes in the Probability of Occurrence of Floods
in Bangladesh and Implications. Global Environmental Change 12 (July): 127–138. https://doi.
org/10.1016/S0959-3780(02)00002-X.
Osgood, Daniel, Bristol Powell, Rahel Diro, Carlos Farah, Markus Enenkel, Molly E. Brown, S.
Greg Husak, Lucille Blakeley, Laura Hoffman, and Jessica L. McCarty. 2018. Farmer Perception,
Recollection, and Remote Sensing in Weather Index Insurance: An Ethiopia Case Study. Remote
Sensing 10 (12): 1887. https://doi.org/10.3390/rs10121887.
Osgood, Daniel, and Kenneth E. Shirley. 2012. “The Value of Information in Index Insurance for
Farmers in Africa.” In The Value of Information: Methodological Frontiers and New Applications
in Environment and Health, edited by Ramanan Laxminarayan and Molly K. Macauley, 1–18.
Dordrecht: Springer Netherlands. https://doi.org/10.1007/978-94-007-4839-2_1.
Pekel, Jean-François., Andrew Cottam, Noel Gorelick, and Alan S. Belward. 2016. High-Resolution
Mapping of Global Surface Water and Its Long-Term Changes. Nature 540 (7633): 418–422.
https://doi.org/10.1038/nature20584.
RMS. 2017. “RMS—Executive Summary.” RMS. https://forms2.rms.com/DFID-Executive-Sum
mary.html.
Robertson, Andrew, Sergey Kirshner, and Padhraic Smyth. 2003. “Hidden Markov Models for
Modeling Daily Rainfall Occurrence over Brazil,” December.
Sabaghy, Sabah, Jeffrey P. Walker, Luigi J. Renzullo, and Thomas J. Jackson. 2018. Spatially
Enhanced Passive Microwave Derived Soil Moisture: Capabilities and Opportunities. Remote
Sensing of Environment 209 (May): 551–580. https://doi.org/10.1016/j.rse.2018.02.065.
Scheel, M. L. M., M. Rohrer, Ch. Huggel, D. Santos Villar, E. Silvestre, and G. J. Huffman. 2011.
“Evaluation of TRMM Multi-Satellite Precipitation Analysis (TMPA) Performance in the Central
Andes Region and Its Dependency on Spatial and Temporal Resolution.” Hydrology and Earth
System Sciences 15 (8): 2649–2663. https://doi.org/10.5194/hess-15-2649-2011.
Tangdamrongsub, Natthachet, Chalita Forgotson, Chandana Gangodagamage, and Joshua
Forgotson. 2021. The Analysis of Using Satellite Soil Moisture Observations for Flood Detection,
Evaluating over the Thailand’s Great Flood of 2011. Natural Hazards. https://doi.org/10.1007/
s11069-021-04804-8.
Tellman, Beth, Upmanu Lall, Pierre Gentine, and Venkat Lakshmi. 2021. “Mapping Flood Impacts
Using Multi-Sensor Satellite Data Fusion in Urban Areas.” NASA Terrestrial Hydrology.
Tellman, Beth, Cody Schank, Bessie Schwarz, Peter D. Howe, and Alex de Sherbinin. 2020. Using
Disaster Outcomes to Validate Components of Social Vulnerability to Floods: Flood Deaths and
Property Damage across the USA. Sustainability 12 (15): 6006. https://doi.org/10.3390/su1215
6006.
Thomas, Mitchell, Elizabeth Tellman, Ben DeVries, Akm Saiful Islam, Michael Steckler, Max
Goodman, and Maruf Billah. in pub. “A Sentinel-1 Inundation Time Series Algorithm for Flood
Index Insurance Applications in Bangladesh.”
ULABY, F. T. 1982. “Microwave Remote Sensing Active and Passive.” Rader Remote Sensing and
Surface Scattering and Emission Theory II: 848–902.
UNISDR. 2015. “The Human Cost of Weather-Related Disasters 1995–2015.” https://www.undrr.
org/publication/human-cost-weather-related-disasters-1995-2015.
United News of Bangladesh. 2020. “WFP, Oxfam, Green Delta Insurance Join Hands to Protect
Labourers,” 2020. https://www.unb.com.bd/category/Bangladesh/wfp-oxfam-green-delta-insura
nce-join-hands-to-protect-labourers/55503.
Winsemius, Hessel C., Brenden Jongman, Ted I. E. Veldkamp, Stephane Hallegatte, Mook Banga-
lore, and Philip J. Ward. 2018. Disaster Risk, Climate Change, and Poverty: Assessing the Global
Exposure of Poor People to Floods and Droughts. Environment and Development Economics 23
(3): 328–348. https://doi.org/10.1017/S1355770X17000444.
Xie, Miao, Zhe Jiang, and Arpan Man Sainju. 2018. “Geographical Hidden Markov Tree for Flood
Extent Mapping.” In Proceedings of the 24th ACM SIGKDD International Conference on Knowl-
edge Discovery & Data Mining, 2545–54. KDD ’18. New York, NY, USA: Association for
Computing Machinery. https://doi.org/10.1145/3219819.3220053.
Zeng, Ziyue, Yanjun Gan, Albert J. Kettner, Qing Yang, Chao Zeng, G. Robert Brakenridge,
and Yang Hong. 2020. “Towards High Resolution Flood Monitoring: An Integrated Method-
ology Using Passive Microwave Brightness Temperatures and Sentinel Synthetic Aperture Radar
Imagery.” Journal of Hydrology 582 (March): 124377. https://doi.org/10.1016/j.jhydrol.2019.
124377.
Rivers, Coasts and Estuaries
Impact of Coriolis Force on the Flow
Field and Sedimentation in Ideally
Shaped Tidal Basins
Nazeat Ameen Iqra, Mohammad Asad Hussain, and M. Shah Alam Khan
Abstract Tidal basins under the Tidal River Management projects in Bangladesh
exhibit complex morphological patterns consisting of substantial spatial and temporal
variability of sedimentation. Sedimentation in the tidal basin is controlled by various
physical and hydraulic factors. Therefore, it is indispensable to understand the domi-
nant factors behind these morphological patterns. Flow field and sedimentation
pattern in Pakhimara Beel (tidal basin), located in southwestern Bangladesh, are
being investigated under the present study. Large-length scales of the beel, shallow
water depth, and low depth-averaged velocity created by the hydraulic gradient of
sediment-laden flows lead to Rossby number (defined as Ro = U/fL) < 1, making the
system susceptible to the effect of Coriolis force. This condition alters the flow field
and, thereby, the sedimentation pattern. Four ideally shaped computation domains
with varying depth, inlet channel location, and orientation have been investigated
through the Delft3D model. The results indicate that, for all domains, the flow
velocity is deflected towards the right in the northern hemisphere due to Coriolis
effect. It is found that at large peripheral areas of the beel with substantially low
flow velocity, the circulations are influenced by Coriolis force, and higher sediment
deposition occurs for higher latitudes. The results imply that the deposition patterns
in large tidal basins are significantly influenced by Coriolis effect.
Keywords Five Coriolis effect · Rossby number · Tidal River Management

(TRM) · Tidal basins · Delft3D
1 Introduction
The Ganges–Brahmaputra-Meghna (GBM) river system carries more than a billion

tons of sediments every year (Gain et al. 2017). Along with the Bay of Bengal, the
coastal zone of Bangladesh is geomorphologically and hydrologically dominated by
the GBM system (Ahmad 2019). Although the land subsidence rate of this region
N. A. Iqra (B) · M. A. Hussain · M. S. A. Khan

214 N. Ameen et al.
is 2–3 mm/year, it is compensated by the large sediment supply (Nicholls et al.

2016). Excluding the Sundarbans mangrove ecosystem, larger part of the southwest
coastal zone of Bangladesh is highly diverse and dynamic (Brammer 2014; Gain
et al. 2017), characterized by an agro-ecological system with extensive tidal rivers,
streams, and tidal basins (Mutahara et al. 2017), and has a potential impact of a rising
sea level due to global warming (Brammer 2014). Since the 1960s, coastal polders
(earthen embankments) have been constructed in the southwest region of the country,
to provide protection from floods and to increase agricultural production (Gain et al.
2017). However, because of the interruption of natural tidal flow to the tidal flood-
plains, siltation has taken place in the adjacent tidal rivers, causing prolonged water-
logging and drainage congestion in the coastal polders (Rashid et al. 2013). As a
result, further sediment accretion process is interrupted, and these polders have lost
1–1.5 m of elevation with respect to mean water level (Auerbach et al. 2015). To
address these complex problems, a temporary restoration of controlled tidal flooding
in polders was introduced conceptually as Tidal River Management (TRM) (van
Staveren et al. 2014). TRM has been considered as an effective approach for sustain-
ably managing the southwest coastal area (Nowreen et al. 2014). The temporary
floodplains inundation process involves: (i) restoring tidal flooding (twice-daily) by
means of temporary and partial removal of embankments (or polders); (ii) scouring
the adjacent river bed; and (iii) depositing sediments within the tidal basins (the
lowest parts of the polders) (van Staveren et al. 2017). Generally, it has been seen
that at the beginning of the operation, the sedimentation was relatively low. Sedi-
ment deposition occurred during the monsoon season compared to the dry season.
Sediment deposition was higher close to the opening of the tidal basin and decreased
gradually to the furthest end of the tidal basins. This uneven sedimentation inside tidal
basins hinders the main purpose of TRM and ultimately created drainage conges-
tion. To tackle the problem, appropriate measures such as compartmentalization or
rotation of openings were required (Gain et al. 2017).
The concept of restoring controlled tidal flooding in basins is inspired by the
projects of the Dutch delta (Islam et al. 2020), involving the lowering or complete
removal of embankment sections, subsequently restoring tidal flood dynamics. But
the significant difference in the dynamic sediment deposition patterns needs to be
adequately understood (van Staveren et al. 2017). Understanding the behaviour of the
tides that cause fluctuating water levels and the in and outflow of water; water facili-
tating the movement of suspended sediment is highly recommended by van Staveren
et al. (2017). In order to achieve the goals of TRM and the effective sediment accu-
mulation process, many studies have been conducted from Bangladesh’s perspective.
Amir et al. (2013) modelled a concept of compartmentalization of the tidal basin with
embankments and observed that the sedimentation spreads uniformly after gradual
removal of embankments along the connecting river. Islam et al. (2020) and Talchab-
hadel et al. (2018) used multiple inlets and flow regulation in their numerical model
to achieve uniform sediment spreading. Also, the relation of inlet cross-sectional
area and tidal volume, which directly indicates the amount of sediment load, was
observed by Shampa and Pramanik (2012) numerically and Talchabhadel et al. (2017)
experimentally.
Impact of Coriolis Force on the Flow Field and Sedimentation … 215
The small depth-average velocity in most parts of the beel, created by the hydraulic
gradient of sediment-laden flows from inlet channels, and large-length scales of the
beels make the influence of Earth’s rotational effect, e.g. the Coriolis effect important
on the flow field and sedimentation pattern (Wells 2007, 2009) in these artificially
compounded water bodies. Wells (2009) used analogue laboratory experiments to
study the influence of Coriolis forces upon the depositional patterns of turbidity
currents. His results showed that when the density of the sediment-laden flow was
kept constant with increasing rotation rates (e.g. latitudes) the length scale decreases,
which means a systematic reduction in the radius within which most of the particles
are deposited. In other words, the Rossby radius of deformation determines the
deposition patterns. Ahn et al. (2016) investigated the influence of Coriolis force
on upstream water flows and showed that the ‘Rossby number’ decreases with the
increase of latitude, keeping inflow velocity and approach length constant (except
at the equator where Rossby becomes infinity). The scale at which Coriolis forces
become important is best expressed by the Rossby number, defined as Ro = U/fL,
where U is a depth-averaged velocity, L the length scale and the Coriolis frequency,
f , is characterized by f = 2wsin θ, where w is the earth’s rotation speed (can be
assumed to be a constant of 7.292 × 10−5 rad/s) and θ is the latitude. Coriolis forces
will become significant when Ro < 1 (Wells 2009; Cossu et al. 2010). When Ro <
1 Coriolis forces will result in a significant deflection of the turbidity current to the
right in the Northern hemisphere (Wells 2009). Cossu et al. (2010) observed that
in the northern hemisphere, the right-hand channel levee being noticeably higher
than the left-hand levee, which indicates the effect of Coriolis forces upon turbidity
currents. The influence of Coriolis in sediment deposition patterns is recognized by
many authors such as Unger and Nemeth (2013). The authors used remote sensing
techniques to estimate the effect of Coriolis force for the Tisza River and found the
right bank is higher than the left bank, and the westward shifting of the river is due
to the impact of Coriolis force. Huijts et al. (2006, 2009) also showed the rightward
deflection of sedimentation influenced by Coriolis force. So, the impacts of Coriolis
force on sedimentation focusing on controlled flooding as in the case of TRM projects
may be an important factor to study the sedimentation patterns inside the beels.
The present study intends to focus on the extent of Coriolis force influencing the
sedimentation process through numerical experiments in idealized model domains.
Idealized models can be a handy tool for studying the influence of various dynamic
factors, keeping the geometry of the channel and other parts of the domain simple
(Maskell et al. 2014). From previous studies, idealized models have been used to
simulate lateral estuarine circulation and the implications for sediment transport and
estuary morphology (Chen and Sanford 2009a), the vertical estuarine structure and
salt transport (Chen and Sanford 2009b), the flood and ebb dominance of sediment
transport (Robins 2008), the dependency of tides and storm surge on cannel depth
(Familkhalili and Talke 2016). Among very few idealized model-based researches
for the assessment of TRM in Bangladesh, a numerical model for solving drainage
congestion (Shampa and Pramanik 2012) and the influence of inlet cross-section
on sedimentation (Talchabhadel et al. 2017) can be found. In the present study,
the effect of the Coriolis force on sediment accumulation is investigated under the
216 N. Ameen et al.
different latitudes conditions, and also the sediment deposition characteristics have
been discussed for selected geometric features of these idealized domains.
2 Study Area
Since 1997, the polders in the southwest region of Bangladesh have been subject to
TRM. After that, many tidal basins went through the TRM operation. Till now, the
timeline of the TRM operation for the different tidal basins (beel) is shown in Table
1 (Talchabhadel et al. 2016).
The study area is Pakhimara Beel (tidal basin), located in the Khulna division
of the southwest (SW) region (Fig. 1). It lies between 22°41 45 N, 89°13 33 E and
22°39 34 N, 89°15 17 E and covers 7 km2 (Gain et al. 2017) having an average slope
of 0.2%. For the TRM project operation, the adjacent Kobadak River is connected
to the beel through a man-made link canal of 1.5 km length and 40 m maximum
bottom width at −1.1 m PWD (the datum for the Public Works Department of
Bangladesh with the zero datum at 0.46 m below MSL is called PWD). The flow
and transport processes of the beel are dominated by tidal dynamics as there is no
regulated gate to control the flow from the river through the canal. The average
tidal ranges during pre-monsoon, monsoon, and dry seasons are 2.4 m, 1.5 m, and
2 m, respectively. On average, the measured Suspended Solid Concentration (SSC)
at Balia on Kobadak River is 0.84 kg/m3 , 0.58 kg/m3 , 0.37 kg/m3 , respectively,
during pre-monsoon, monsoon, and dry season (Islam et al. 2020). Cohesive particles
dominate the suspended sediments as the median grain size is 64 μm. Maximum
sediment concentration was found during April.
Before TRM, the beel had a mean elevation of 0.57 m PWD. After 1.5 years,
a survey was conducted to assess the sediment deposition progress inside the beel.
The mean elevation was found to be 1.17 m PWD. Though the land level increased
due to sediment deposition, a significantly uneven distribution of sediments was
found. Sediments mostly deposited close to the inlet mouth, and the elevation at the
peripheral portions of the beel remained unchanged (Islam et al. 2020).
Table 1 The timeline of the TRM in Bangladesh

1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
Beel Bhaina
Beel Kedaria
Beel Khuksia
Beel Kapalia
Beel Pakhimara*
*The study area

Fig. 1 Map of the study area
3 Methodology
A depth-averaged two-dimensional Delft3D numerical model has been configured

with idealized geometry and boundary conditions which are derived from the field
conditions. Delft3D-Flow (Delft3D-FLOW 2011) model is a multidimensional (2D
or 3D) open-source hydrodynamic and sediment transport model solving the unsteady
shallow water equations. It is a computational tool for coastal, river, and estuarine
areas which can simulate flow, sediment transport, wave, and morphological develop-
ments. A rectangular grid (Cartesian frame of reference) is considered as a simplified
218 N. Ameen et al.
form of a curvilinear grid. The total number of grid cells in the computation domain
is 61468. The size of grid cells is smaller in the river and connecting channel, and the
size of grid cells increases inside the beel. The surface area of the grid cells varies
from 24 m2 to 660 m2 . From the measured water level data in the study area, it is
found that the tidal range in the Kobadak River is 2.6 m. A sinusoidal tidal water
level variation (ignoring tidal asymmetry) is used as downstream boundary keeping
tidal range 2.6 m resembling the study area. Keeping similarity with observed data,
relatively low river discharge (varying from 25 ~ 70 m3 /s) is used as an upstream
boundary as the model is simulated for 30 days of the dry period. The applied water
level and discharge data for 24 h at the model boundary are shown in Fig. 2.
For the model domain, an idealized bathymetry is used, keeping the maximum
depth of 3.5 m for the beel. The maximum depth of the canal and the river is 3.5 m
and the width are 30 m and 50 m, respectively. The sediment size of the coastal rivers
was studied by Mahiuddin (2017) and they found that silt content in the suspended
sediment is more than 50% and sand content ranged from 16 to 24%, while clay
content ranged from 18 to 23%. For the present study, sand is considered to be 24%
as non-cohesive soil, and silt and clay to be 76% as cohesive soil. Suspended sediment
concentration (=0.84 kg/m3 ) (Islam et al. 2020) is divided into this same percentage
for the present research. The applied median sediment diameter, d 50 is equal to 64 μm.
The critical bed shear stress of the model for sedimentation and erosion is 1000 N/m2
and 0.01 N/m2 , respectively. The manning roughness coefficient 0.03 s/m1/3 was used
for this model study.
In the present research, several ideally shaped domains are constructed resembling
the study area to investigate the impact of beel geometry on flow field and sedimen-
tation (Fig. 3). To understand the impact on the flow field and sedimentation, the
model domains were run for two latitudes: 22.7° N and 0° N. Four ideally shaped
computation domains with varying orientation and depth have been subjected to the
investigation (Fig. 3). Scenario (i) is a rectangular-shaped domain with a constant
80 1.5
70
1
60
Discharge, m3/s
0.5
Water Level (m)
50
40 0
30
-0.5
20
-1
10
0 -1.5
21:36 0:00 2:24 4:48 7:12 9:36 12:00 14:24 16:48 19:12 21:36 0:00
Time, hr
Discharge Data Water Level data
Fig. 2 The model upstream boundary discharge and downstream boundary tidal water level data
Scenario (i) Scenario (ii)
Elevaon (m)
Inflow Inflow
Scenario (iii) Scenario (iv)
Inflow Inflow
Fig. 3 Four scenarios of the idealized shaped domains

220 N. Ameen et al.
depth of 3.5 m; Scenario (ii) is the same domain tilted towards the left forming the
similar orientation of the study area keeping the previous depth; Scenarios (iii) and
(iv) the model domains are further extended towards north connected with the titled
domain of Scenario (ii). The domain for Scenario (iii) has a constant depth while that
for Scenario (iv) partially consists of a sloped bottom. Due to the unavailability of
the detailed bathymetry of the study area, the concept of an ideal-shaped domain is
used. Also, the main objective of the study is to identify the influence of the physical
factors such as Coriolis force on the sediment deposition in the beel. Therefore, the
ideal-shaped domain has been experimented numerically. Although the Scenarios
(iii) and (iv) ideally represent the actual beel, before analysing them, the simpler
shape in Scenarios (i) and (ii) was considered. The concept of sequential change
of domains is found to be helpful to identify the sedimentation in terms of shape,
orientation, and bathymetry. Also, for all the scenarios and in both latitudes, the
model parameters including boundary conditions and physical parameters are kept
the same only to focus on the impact of Coriolis force.
4 Results and discussion
4.1 Testing the Coriolis Effect in the Study Area
At first, to examine whether the flow field of the beel in the study area (at 22.7° N
latitude) is influenced by the Coriolis effect, the Rossby number (Ro) is calculated
for a range of depth-average velocities and beel sizes (representing length scales).
Table 2 shows the calculated Rossby numbers at 22.7° N latitude for depth-average
velocities 1 ~ 20 cm/s and length scales of 0.5 ~ 5 km. Values with Ro < 1 are shaded
in the table. Calculated values show that for a length scale of 2 km, Rossby numbers
Ro remain less than unity for depth-average velocity of up to 10 cm/sec.
As already mentioned, whether the Coriolis force should be taken into account or
not depends on the Rossby number to be equal or less than 1. Rossby number (Ro
= U/fL) is dependent on the flow velocity, Coriolis frequency, and length. In Table
2, it is shown that the Rossby number varies with velocity and beel size at a fixed
latitude. We assume our domain is under Coriolis influence. So, the max value of
Table 2 Changes of Rossby number with depth-average velocity and length scale at a given latitude
Depth average Beel size (km)
velocity (cm/s) 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
1 0.36 0.18 0.12 0.09 0.07 0.06 0.05 0.04 0.04 0.04
2 0.71 0.36 0.24 0.18 0.14 0.12 0.10 0.09 0.08 0.07
5 1.78 0.89 0.59 0.44 0.36 0.30 0.25 0.22 0.20 0.18
10 3.55 1.78 1.18 0.89 0.71 0.59 0.51 0.44 0.39 0.36
15 5.33 2.66 1.78 1.33 1.07 0.89 0.76 0.67 0.59 0.53
20 7.10 3.55 2.37 1.78 1.42 1.18 1.01 0.89 0.79 0.71
Rossby number, Ro has to be equal to one. At a certain latitude when Ro = 1, from

Rossby formula, we get that the length of influence is proportional to the velocity
(L ~ U). Based on this concept, the length scale (L) is measured for each scenario.
The velocity will decrease as the flow spreads across the domain. At the mouth of
the basin, the flow velocity will be high but the velocity will drop tremendously after
crossing a certain distance from the mouth of the canal.
Figure 4 shows the length scale of influence calculated for each domain keeping
latitude at 22.7 degrees. Figure 4 shows that for Scenario (i), after 1300 m, the Coriolis
force should be taken into account. For Scenario (ii), the value is 1200 m, and for
Scenario (iii) and Scenario (iv), the value is 1000 m which means that if the domain
is bigger, the Coriolis influence zone is larger. So, at the mouth of the link canal,
the inertia force dominates the flow and Coriolis force can be neglected. And after
a certain distance, the Coriolis force also contributes to the flow field and thereby
sediment distribution.
Scenario (i) Scenario (ii) Length

of
influence
(m)
Scenario (iii) Scenario (iv)
Fig. 4 The length scale variation for the four scenarios

222 N. Ameen et al.
5 Results from Numerical Experiments
For all the numerical experiments performed during the present study, the forcing
factors like downstream tidal water level, upstream river discharge, and friction
factors were kept constant. This ensures that when the latitude is altered i.e. the
Coriolis coefficient is changed, the modified flow field and thereby the change in
sedimentation are only caused by the Coriolis force. We anticipate that the change in
the flow field and sedimentation caused by the change in Coriolis force will be very
small but in field conditions, it may play an important role in combination with all
other forcing factors.
At first, spatial distribution of depth-average velocities during peak flood for the
four computation domains are presented in Fig. 5, where the left panel shows the
depth-average velocities at 0° latitude and the right panel shows the same for 22.7°
latitude.
From Fig. 5, it is evident that at the mouth of the link canal, jet-flow with flow
separation phenomenon appears due to high inertia force for all the domains at both
latitudes. But at locations \away from the mouth of the link canal depth, average flow
velocities are less than 10 cm/s even during peak flood periods. So, it can be expected
that the flow field in the areas away from the mouth of the link canal is influenced
by the Coriolis force. From the same figure, with the change of beels’ orientation
and depth, the change in spatial extent of the zone where the influence of Coriolis
force is expected to be significant can be clearly identified. For Scenarios (i) and (ii),
the shape of the jet-flow area appears to be elliptical shape, whereas for Scenarios
(iii) and (iv), they are circular. This is same for both the latitudes. But the spatial
distribution for depth-average velocities at the peripheral areas show some difference
between 0° latitude (left panel) and 22.7° latitude (right panel). This difference must
be due to the impact of the Coriolis effect as the other forcing forces have been
kept constant. To observe the difference in depth-average velocities in the peripheral
areas, cross-sections across the peripheral areas are plotted. Figure 6 shows the cross-
sectional view of the depth-average velocities for four scenarios comparing the two
latitudes.
The difference in depth-average velocities in the peripheral areas are clearly visible
from the cross-sectional profiles (for the northern part, 1500 m away, and for the
southern part, 1200 m away from the mouth of the canal) of Fig. 6. Especially for
Scenarios (iii) and (iv), it is clear that at higher latitude, depth-average velocity is
higher and the difference is prominent in the northern part. This phenomenon might
be due to the fact that the flow entering into the beel through the link canal is being
deflected towards the right by the Coriolis force. As a result, at higher latitude where
the Coriolis effect is larger, the depth-average velocity becomes significantly higher
in the northern part.
Next, the spatial variation of accumulated sedimentation after 30 days of model
runs for Scenarios (i), (ii), (iii), and (iv) at 0° latitude (left panel) and 22.7° latitude
(right panel). Similar to the flow field, it is evident that at the mouth of the link canal,
significant accumulation of sediment occurs for all the domains at both latitudes.
Depth average
velocity
Scenario (i)
(m/s)
i
i
Scenario (ii)
Scenario (iii)
i
Scenario (iv)
Fig. 5 Spatial distribution of depth-average velocities during peak flood for Scenarios (i), (ii), (iii),
and (iv) at 0° latitude (left panel) and 22.7° latitude (right panel)
224 N. Ameen et al.
velocity scenario (i) velocity scenario (i)

(m/s) 0 degree 22.7 degree (m/s) 0 degree 22.7 degree
1.1E-02 3.3E-02
2.8E-02
9.0E-03
2.3E-02
7.0E-03 1.8E-02
1.3E-02
5.0E-03
8.0E-03
3.0E-03 3.0E-03
0 500 1000 1500 0 500 1000 1500
distance (m) from west to east distance (m) from west to east
velocity scenario (ii) velocity scenario (ii)
(m/s) 0 degree 22 degree (m/s) 0 degree 22.7 degree
3.1E-03 1.3E-03
1.1E-03
2.1E-03 9.0E-04
7.0E-04
1.1E-03 5.0E-04
3.0E-04
1.0E-04 1.0E-04
0 100 200 300 400 500 600 500 700 900
velocity scenario (iii) velocity scenario (iii)
(m/s) 0 degree 22.7 degree
(m/s) 0 degree 22.7 degree
5.0E-03 1.7E-02
4.0E-03 1.2E-02
3.0E-03 7.0E-03
2.0E-03 2.0E-03
0 100 200 300 400 500 600 700 1300 1500 1700 1900
velocity scenario (iv) velocity scenario (iv)

(m/s) 0 degree 22.7 degree (m/s) 0 degree 22.7 degree
5.0E-03 1.7E-02
4.0E-03 1.2E-02
3.0E-03 7.0E-03
2.0E-03 2.0E-03
0 100 200 300 400 500 600 700 1300 1500 1700 1900
Fig. 6 Cross-sectional profiles of depth-average velocities for two latitudes at the northern part
(left panel) and southern part (right panel)
But at locations away from the mouth of the link canal, amount of sedimentation is
very small even after the same time durations. The flow field in these areas which
is away from the mouth of the link canal is influenced by the Coriolis force. The
spatial distribution for sedimentation at the peripheral areas shows some difference
between 0° latitude (left panel) and 22.7° latitude (right panel).
Next, the spatial distribution of sedimentation of the four domains and cross-
sectional profiles for sedimentation at the northern and southern peripheral areas of
the domains are plotted to observe the difference in sedimentation due to the change
in latitude. Figure 7 shows the spatial distribution and Figure 8 shows the cross-
sectional view of the sedimentation for four scenarios comparing the two latitudes.
The significant difference of sediment deposition at two latitudes cannot be identified
from the spatial map in Fig. 7. However, uneven sediment distribution in the domains
is noticeable. Also, along the domain peripheral, it is evident that the sediment
accumulation varies with the depth variation. Such as for Scenario (iv), along the
southwest peripheral, the deposition is lower than for Scenario (iii).
The difference in sedimentation in the northern peripheral areas is clearly visible
from the cross-sectional profiles of Fig. 8 (northern part). It is clear that at higher
latitude, sedimentation is higher at the northern parts of the beels. Like depth-average
velocity distributions, this phenomenon again might be due to the fact that the flow
entering into the beel through the link canal is being deflected towards the right by
the Coriolis force, and as a result, higher sediment is being deposited in the northern
peripheral areas of the beels. So, at higher latitude where the Coriolis effect is larger,
sedimentation becomes significantly higher in the northern part.
According to (Wells 2009), for higher latitude, one will expect a smaller radius
of sedimentation and a higher deflection of sedimentation towards the right. Thus,
with the increase of latitude, the deposition decreases at a certain radius and increases
towards the right. From Fig. 8 (southern part), it can be seen that for 0°, the deposition
is higher than 22.7°. This indicates that the radius of sedimentation decreases with
increasing latitude. This is because at a lower latitude, due to smaller particle size,
the settling velocity increases. This makes the turbidity current weaker and sediment
spreads in more area. The d 50 of the sediment particle is 64 μm with a settling
velocity of 0.0025 m/s, which is small enough to keep the sediment in suspension,
making the flow velocity slower to spread more under the influence of earth rotation.
Also, in Fig. 8 (southern part) for Scenario (i), the deposition increases from west
to east, whereas for Scenarios (ii), (iii), and (iv), the deposition decreases. This is
due to the orientation of the domains of Scenarios (ii), (iii), (iv) such that the flow
deflecting towards the right carrying most of the sediment. As a result, the deposition
in the southeast part of the domains of Scenarios (ii), (iii), and (iv) is lower than in
the southwest.
226 N. Ameen et al.
accumulaon
Scenario (i)
(m)
Scenario (ii)
i
Scenario (iii)
Scenario (iv)
Fig. 7 Spatial distribution of accumulated sediment deposition after 30 days of model runs for
Scenarios (i), (ii), (iii), and (iv) at 0° latitude (left panel) and 22.7° latitude (right panel)
scenario (i) scenario (i)

0 degree 22.7 degree 0 degree 22.7 degree
1.1E-03 7.5E-03
accumulaon (m)
accumulaon (m)
6.5E-03
1.0E-03
5.5E-03
9.0E-04 4.5E-03
500 700 900 1100 1300 1500 500 1000 1500
scenario (ii) scenario (ii)

0 degree 22 degree 0 degree 22.7 degree
1.6E-03 1.2E-02
1.5E-03 1.1E-02
accumulaon (m)
accumulaon (m)
1.4E-03 1.1E-02
1.3E-03 1.0E-02
1.2E-03 9.5E-03
1.1E-03 9.0E-03
100 200 300 400 500 500 700 900
scenario (iii) scenario (iii)

1.6E-03 1.26E-02
1.5E-03
accumulaon (m)
accumulaon (m)
1.4E-03 1.21E-02
1.3E-03
1.2E-03 1.16E-02
1.1E-03
1.0E-03 1.11E-02
0 100 200 300 400 500 1500 1600 1700 1800 1900
scenario (iv) scenario (iv)

1.6E-03 1.23E-02
1.5E-03
accumulaon (m)
accumulaon (m)
1.4E-03 1.18E-02
1.3E-03
1.2E-03 1.13E-02
1.1E-03
1.0E-03 1.08E-02
0 100 200 300 400 500 1500 1600 1700 1800 1900
Fig. 8 Cross-sectional profiles of sedimentation for two latitudes at the northern part (left panel)
and southern part (right panel)
228 N. Ameen et al.
6 Conclusions
Even and adequate sedimentation in the beel is the most necessitated objective to be
achieved for an effective and successful TRM operation. So, determining the extent
of influence of physical and hydraulic factors controlling the sedimentation process
is indispensable. The impact of Coriolis force on sedimentation is minor yet hard
to neglect. This study shows how Coriolis force impacts the spatial distribution of
sedimentation at a different latitude. The Rossby number is a very helpful parameter
to determine the influence of Coriolis force in the study area. The ‘radius of defor-
mation’ which is expressed as length scale in this study can be determined based on
Rossby number equal to one. In this study, the spatial sediment distribution in an
idealized domain at the two different latitudes shows that the depth of sedimentation
is more in higher latitude and it is deflected towards the right side of the inlet while
the opposite happens at the left side of the domain. The Coriolis force is negligible
at the mouth of the link canal where the velocity is mostly dominated by the inertia
force. After a certain length (=1000 m on average), the Coriolis force dominates. The
authors believe that this study will add another perspective to focus on the causes
of uneven sedimentation in the tidal basin in terms of TRM and how Coriolis force
incorporates with other physical and hydraulic factors.
Acknowledgements The research was funded by the ‘Living Polders’ project of the Institute of
Water and Flood Management (IWFM), Bangladesh University of Engineering and Technology
(BUET), Bangladesh.
References
Ahmad, Hafez. 2019. Bangladesh Coastal Zone Management Status and Future Trends. Journal of
Coastal Zone Management 22 (1): 1–7. https://doi.org/10.4172/2473-3350.1000466.
Ahn, S.H., Y.X. Xiao, X.Z. Zhou, J. Zhang, C.J. Zeng, Y.Y. Luo, W. Xu, and Z.W. Wang.
2016. Numerical Analysis of Coriolis Effect on Low-Head Hydraulic Turbines. IOP Conference
Series: Earth and Environmental Science 49 (2): 022012. https://doi.org/10.1088/1755-1315/49/
2/022012.
Amir, Md. Sharif Imam Ibne, M. Shah Alam Khan, Mohammad Masud Kamal Khan, Mohammad
Golam Rasul, and Fatema Akram. 2013. Tidal River Sediment Management—A Case Study in
Southwestern Bangladesh. International Journal of Geological and Environmental Engineering
7 (3): 175–185.
Auerbach, L.W., S.L. Goodbred Jr., D.R. Mondal, C.A. Wilson, K.R. Ahmed, K. Roy, M.S. Steckler,
C. Small, J.M. Gilligan, and B.A. Ackerly. 2015. Flood Risk of Natural and Embanked Landscapes
on the Ganges-Brahmaputra Tidal Delta Plain. Nature Climate Change 5 (2): 153–157. https://
doi.org/10.1038/nclimate2472.
Brammer, Hugh. 2014. Bangladesh’s Dynamic Coastal Regions and Sea-Level Rise. Climate Risk
Management 1: 51–62. https://doi.org/10.1016/j.crm.2013.10.001.
Chen, Shih-Nan, and Lawrence P. Sanford. 2009a. Lateral Circulation Driven by Boundary Mixing
and the Associated Transport of Sediments in Idealized Partially Mixed Estuaries. Continental
Shelf Research 29 (1): 101–118. https://doi.org/10.1016/j.csr.2008.01.001.
Chen, Shih-Nan, and Lawrence P. Sanford. 2009b. Axial Wind Effects on Stratification and Longitu-
dinal Salt Transport in an Idealized, Partially Mixed Estuary. Journal of Physical Oceanography
39 (8): 1905–1920. https://doi.org/10.1175/2009jpo4016.1.
Cossu, R., M.G. Wells, and A.K. Wåhlin. 2010. Influence of the Coriolis Force on the Velocity
Structure of Gravity Currents in Straight Submarine Channel Systems. Journal of Geophysical
Research 115 (C11). https://doi.org/10.1029/2010jc006208.
Delft3D-FLOW, U.M. 2011. WL| Delft Hydraulics, the Netherlands.
Familkhalili, R., and S.A. Talke. 2016. The Effect of Channel Deepening on Tides and Storm Surge:
A Case Study of Wilmington, NC. Geophysical Research Letters 43 (17): 9138–9147. https://
doi.org/10.1002/2016gl069494.
Gain, Animesh K., David Benson, Rezaur Rahman, Dilip Kumar Datta, and Josselin J. Rouillard.
2017. Tidal River Management in the South West Ganges-Brahmaputra Delta in Bangladesh:
Moving Towards a Transdisciplinary Approach? Environmental Science & Policy 75: 111–120.
https://doi.org/10.1016/j.envsci.2017.05.020.
Huijts, K. M. H., H. M. Schuttelaars, H. E. de Swart, and A. Valle-Levinson. 2006. Lateral entrap-
ment of sediment in tidal estuaries: An idealized model study. Journal of Geophysical Research
111 (C12). https://doi.org/10.1029/2006jc003615.
Huijts, K. M. H., H. M. Schuttelaars, H. E. de Swart, and C.T. Friedrichs. 2009. Analytical study
of the transverse distribution of along-channel and transverse residual flows in tidal estuaries.
Continental Shelf Research 29 (1): 89–100. https://doi.org/10.1016/j.csr.2007.09.007.
Islam, Md. Feroz, Hans Middelkoop, Paul P. Schot, Stefan C. Dekker, and Jasper Griffioen.
2020. Enhancing Effectiveness of Tidal River Management in Southwest Bangladesh Polders by
Improving Sedimentation and Shortening Inundation Time. Journal of Hydrology 590: 125–228.
https://doi.org/10.1016/j.jhydrol.2020.125228.
Mahiuddin, Md. 2017. Siltation Pattern and Pedogenesis in the Coastal Zone of Bangladesh. PhD
Dissertation, University of Dhaka.
Maskell, J., K. Horsburgh, M. Lewis, and P. Bates. 2014. Investigating river–surge interaction in
idealised estuaries. Journal of Coastal Research 294: 248–259. https://doi.org/10.2112/jcoastres-
d-12-00221.1.
Mutahara, Mahmuda, Jeroen F. Warner, Arjen E. J. Wals, M. Shah Alam Khan, and Philippus
Wester. 2017. Social Learning for Adaptive Delta Management: Tidal River Management in
the Bangladesh Delta. International Journal of Water Resources Development 34 (6): 923–943.
https://doi.org/10.1080/07900627.2017.1326880.
Nicholls, R.J., C.W. Hutton, A.N. Lázár, A. Allan, W.N. Adger, H. Adams, J. Wolf, M. Rahman, and
M. Salehin. 2016. Integrated Assessment of Social and Environmental Sustainability Dynamics
in the Ganges-Brahmaputra-Meghna Delta, Bangladesh. Estuarine, Coastal and Shelf Science
183: 370–381. https://doi.org/10.1016/j.ecss.2016.08.017.
Nowreen, Sara, Mohammad Rashed Jalal, and M. Shah Alam Khan. 2014. Historical Analysis of
Rationalizing South West Coastal Polders of Bangladesh. Water Policy 16 (2): 264–279. https://
doi.org/10.2166/wp.2013.172.
Rashid, Md. Bazlar, Arif Mahmud, Md. Kamrul Ahsan, Md. Hossain Khasru, and Md. Ashraful
Islam. 2013. Drainage Congestion and Its Impact on Environment in the South-Western Coastal
Part of Bangladesh. Bangladesh Journal of Geology 26: 359–371.
Robins, Peter. 2008. Present and Future Flooding Scenarios in the Dyfi Estuary, Wales, U.K.
Countryside Council for Wale.
Shampa, and Md. Ibne Mayaz Pramanik. 2012. Tidal River Management (TRM) for Selected Coastal
Area of Bangladesh to Mitigate Drainage Congestion. International Journal of Scientific &
Technology Research 1 (5): 1–6.
Staveren, Martijn F. van, Jeroen F. Warner, and M. Shah Alam Khan. 2017. Bringing in the Tides.
From Closing Down to Opening Up Delta Polders Via Tidal River Management in the Southwest
Delta of Bangladesh. Water Policy 19 (1): 147–164. https://doi.org/10.2166/wp.2016.029.
230 N. Ameen et al.
Staveren, Martijn F. van, Jeroen F. Warner, Jan P.M. van Tatenhove, and Philippus Wester. 2014.
Let’s Bring in the Floods: De-Poldering in the Netherlands as a Strategy for Long-Term Delta
Survival? Water International 39 (5): 686–700. https://doi.org/10.1080/02508060.2014.957510.
Talchabhadel, Rocky, Hajime Nakagawa, and Kenji Kawaike. 2018. Sediment Management in Tidal
River: A Case Study of East Beel Khuksia, Bangladesh. In ed. A. Paquier and N. Rivière. E3S
Web of Conferences 40: 02050. https://doi.org/10.1051/e3sconf/20184002050.
Talchabhadel, Rocky, Hajime Nakagawa, Kenji Kawaike, Masakazu Hashimoto, and Nassim
Sahboun. 2017. Experimental Investigation on Opening Size of Tidal Basin Management: A
Case Study in Southwestern Bangladesh. Journal of Japan Society of Civil Engineers, Ser. B1
(Hydraulic Engineering) 73 (4): I_781–1_786. https://doi.org/10.2208/jscejhe.73.i_781.
Talchabhadel, Rocky, Hajime Nakagawa, and Kenji Kawaike. 2016. Experimental Study on
Suspended Sediment Transport to Represent Tidal Basin Management. Journal of Japan Society
of Civil Engineers, Ser. B1 (Hydraulic Engineering) 72 (4): I_847–1_852. https://doi.org/10.
2208/jscejhe.72.i_847.
Unger, Zoltan, and Nemeth Istvan. 2013. Estimation of the Coriolis-Force Size for Tisza River.
Carpathian Journal of Earth and Environmental Sciences 8 (2): 201–208.
Wells, M.G. 2007. Influence of Coriolis Forces on Turbidity Currents and Sediment Deposi-
tion. Particle-Laden Flow, ERCOFTAC Series 11: 331–343. https://doi.org/10.1007/978-1-4020-
6218-6_26.
Wells, M.G. 2009. How Coriolis Forces Can Limit the Spatial Extent of Sediment Deposition of a
Large-Scale Turbidity Current. Sedimentary Geology 218 (1–4): 1–5. https://doi.org/10.1016/j.
sedgeo.2009.04.011.
The Impact of Small Tributaries Flood
in the Braided Plain of Large River
Md. Manjurul Hussain, Shampa, Juwel Islam, Md. Shibbir Ahmed,

Md. Ashiqur Rahman, and Md. Munsur Rahman
Abstract Flooding of the braided Brahmaputra-Jamuna River causes significant

economic losses in the northern region of Bangladesh every year. Presently, the
intensity and timing of floods are mainly captured by the water level at a few
selected stations of the river. Flood Forecasting and Warning Centre (FFWC) of
the Bangladesh Water Development Board (BWDB) is officially mandated to regu-
larly disseminate the water level for several rivers during the flood season. The water
levels along the Brahmaputra-Jamuna at some strategic stations (such as Chilmari,
Bahadurabad, Sirajganj etc.) are playing an important role for the above flood
warning system in the North-Central part of Bangladesh. The right bank tributaries to
the Brahmaputra-Jamuna River—the Teesta and the Dharla—are contributing rapid
(flashy in nature) flow to the system, the effects of which are still unidentified. This
study investigates the effects of the variations of flood flow of these tributaries on
the flooding of the Brahmaputra-Jamuna basin at the local level by using the 2D
hydrodynamic model. A 225 km long hydrodynamic model domain was set along
the Brahmaputra-Jamuna, including the most vulnerable floodplain of Dharla and
Teesta at Kurigram and Jamalpur districts, respectively. The results show that the
variation of flow of the tributaries is trivial to the flood level at the stations where
BWDB is capturing water-level data. However, the tributaries’ flow has consider-
able effects on Charland flooding. In addition, the results demonstrate that the flood
depth of some places in the char lands varies more than one metre for the variation of
the contribution it receives from the tributaries. These observations indicate that the
above fact for flood forecasting and warning systems will be important, especially
for the Charland area in the Brahmaputra-Jamuna system.
Keywords Brahmaputra-Jamuna · Dharla · Teesta · Tributaries · Charland ·

Flooding
Md. M. Hussain (B) · Shampa · J. Islam · Md. S. Ahmed · Md. A. Rahman · Md. M. Rahman
232 M. M. Hussain et al.
1 Introduction
Bangladesh is one of the world’s most natural disaster-prone country where floods,
storm surges, cyclones, riverbank erosions and droughts are frequent occurrences
(Nasreen 2004). Among these natural disasters, flood represents one of the most
dominant ones. Flood is a regular feature in Bangladesh due to its geographical setting
influenced by the overflow of major rivers and their tributaries and distributaries
(Karim and Mimura 2008; Mamun et al. 2019; Younus et al. 2007). Catastrophic
flooding events, for example, floods from the years 1988, 1998 and 2020, are a highly
critical issue for the country which causes untold suffering to millions of people. The
effect of catastrophic floods can be devastating and result in appreciable damage to
crops and houses, severe riverbank erosion with consequent loss of homesteads,
schools and land, as well as loss of human lives, livestock and fisheries (Pramanik
1991). If we consider basin-wise flooding, the mighty Brahmaputra-Jamuna has been
the main contributor to most of the catastrophic flood cases in Bangladesh (NAWG
2020).
The Brahmaputra-Jamuna River, draining from the northern and eastern slopes of
the Himalayas, is 2900 km long, wherein the reach length is 240 km in Bangladesh
(Bhuiyan 2014). Geologically, it is one of the youngest rivers in the world (Archana
et al. 2012). Because of abnormal flood and tectonic activity, the Brahmaputra River
began to flow through a new course known as Jamuna from 1787 (Uddin et al. 2011).
The right bank of the Jamuna was once a part of the Teesta floodplain, and now
through the Dudhkumar and Dharala distributary of the Jamuna, is part of the bigger
floodplain. Several distributaries of the Brahmaputra-Jamuna flow through the left
bank floodplain which is later sub-classed as the Brahmaputra-Jamuna floodplain.
The southern part of this sub-region was once a part of the Ganges floodplain (Paul
2011). The Brahmaputra-Jamuna River contributes ∼51% of the water discharge
and 38% of the sediment yield to the Ganges–Brahmaputra-Meghna (GBM) system
(Schumm and Winkley 1994). The river is braided in nature with a width of up to
15 km and average depths of 6 m to 7 m (Klaassen and Vermeer 1988; Shampa
2019a). It has three main tributaries in the left bank i.e. Dudhkumer, Dharla and
Teesta, and three main distributaries in the right bank which are Old Brahmaputra,
Jenai and Dhaleswari (see Fig. 1).
The transboundary Teesta River is one of the major contributors or tributaries of
Brahmaputra-Jamuna. The river flows through five northern districts of Bangladesh,
which are Gaibandha, Kurigram, Lalmonirhat, Nilphamari and Rangpur (Rangpur
Division), comprising an area of 9667 sq.km, 35 thanas and 5427 villages with
an estimated population of 9.15 million as of 2011. The river plays a key role in
flushing silt and sediment deposited during the dry season is a lifeline for irrigation,
agriculture, farming, fishing and navigation in the region. This transboundary river
is encroached by many dams and barrages both in India and Bangladesh. In the rainy
season, excessive flow is dispensed by these dams and barrages, causing floods in the
downstream areas (Rahman and Ali 2016). The river Teesta experiences several peaks
during the monsoon season causing local floods. For example, in the year 2017, the
The Impact of Small Tributaries Flood in the Braided … 233
Fig. 1 The Brahmaputra-Jamuna River and its major tributaries and distributaries
water level crossed the danger level at Dhalia Point of Teesta for more than 6 days, a
resulting in low-lying areas being submerged for several days (FFWC 2018). Among
the other tributaries, Dharla is one of the major contributors to Brahmaputra-Jamuna.
It originates in the Himalayas and enters Bangladesh through the Lalmonirhat district
and flows as the Dharla River until it meets the Brahmaputra-Jamuna River near
the Kurigram District (Fig. 1). Dharla River, along with the Brahmaputra River,
has substantial influence on floods and riverbank erosion in different districts of
Bangladesh (Pal et al. 2017). For example, the water level of the Dharla river at
Kurigram point remained above the danger level for more than 30 days due to the
1998 floods, as a result of which the area was inundated for a long time (FFWC
2018). While the overall flooding situations of past events (i.e. 1987, 1988, 1998,
2007) are quite well documented, the contribution of these two major tributaries in
the flooding of the Brahmaputra-Jamuna is yet to be known (Best et al. 2007; Dewan
et al. 2003; Islam and Chowdhury 2002; Mosselman 2006).
Nevertheless, human activities such as land use and construction of infrastructure,
such as dams, barrages, bridges, artificial levees, are continuously altering the river
regime. Given the large-scale approach, it is important to understand the sub-regions
scale details of the flooding characteristics. In recent decades, flood damage has been
increasing all over the world, due to high population growth, economic activities in
floodplain areas, and intensified precipitation due to climate change (Albano et al.
2017; Kvočka et al. 2016; Munich Re 2015). Bangladesh is no exception (Dewan
2015). Although the country has been able to significantly reduce the loss of lives due
to floods, the economic loss and damage have increased over the decades (Ferdous
et al. 2019; Mechler and Bouwer 2015). The recent monsoon floods of 2020 affected
about 30 districts in Bangladesh, of which 15 districts were severely hit. The damage
was comparatively higher in the Brahmaputra-Jamuna floodplain areas of the Kuri-
gram and Jamalpur districts. More than 50% of the area in 59 unions of 7 Upazilas
of Jamalpur district , and 59 unions of 9 Upazilas of Kurigram district, were flooded
(NAWG 2020).
Against this backdrop, this study attempts to deduce the impact of flood flow of
Teesta and Dharla tributaries on the river basin of Brahmaputra-Jamuna including
its char areas. Specifically, the study attempts to identify the impacts of tributary
flow on flood hazard parameters, such as flood depth, duration and velocity, of the
major Brahmaputra-Jamuna River using 2D hydrodynamic simulations. The flooding
condition of the year 1998 has been considered as the base, and three hypothetical
conditions have been generated based on the flow of Dharla and Teesta. As a study
area, the major flood-prone unions of Kurigram and Jamalpur districts were consid-
ered as shown in Fig. 1. The next section of the paper describes the methodology of
this study. The impact of the tributary flood is illustrated in the last parts.
2 Methodology
For this research, we used a well-calibrated and validated 2D hydrodynamic model by

Shampa (2019b) and Shampa et al. (2017). Delft3D (flow version 4.00.01.000000), an
open-source software, was used to develop the numerical model (Lesser et al. 2004).
In the hydrodynamic part, Navier–Stokes-based two-dimensional depth-averaged
shallSow water (for incompressible free surface flow) equations with consideration
of Boussinesq approximations were used to solve the model. A short description of
the model is given below.
The continuity equation was used to calculate conservation of mass (1)
∂h ∂(hu) ∂(hv)
+ + =0 (1)
∂t ∂x ∂y
Conservation of momentum in the x-direction is shown by Eq. (2)

2
∂u ∂u ∂u ∂ζ gn 2 u u 2 + v 2 ∂ u ∂ 2u
+u +v +g +√ − v h + =0 (2)
∂t ∂x ∂y ∂x 3
h h ∂x2 ∂ y2
Conservation of momentum in the y-direction is shown by Eq. (3)

2
∂v ∂v ∂v ∂ζ gn 2 v u 2 + v 2 ∂ v ∂ 2v
+u +v +g +√ − vh + 2 =0 (3)
∂t ∂x ∂x ∂x 3
h h ∂x2 ∂y
where ζ is water level elevation with respect to a datum (here in m); h repre-
sents water depth (m) u, v is depth average velocity in the x and y directions
(m/s); g is the gravitational acceleration (m/s2 ); vh denotes kinetic eddy
2 viscosity
(m /s); n represents the Manning’s coefficient (sm-1/3). The terms vh ∂∂ xu2 + ∂∂ yu2
2
2
2
and vh ∂∂ xv2 + ∂∂ yv2 in Eqs. (2) and (3) represent the horizontal Reynold’s stress
2
under the eddy viscosity concept neglecting the shear stress along the closed bound-
aries. The model area consists of the whole Brahmaputra-Jamuna River and parts
of Dudhkumar, Dharla, Teesra and Old Brahmaputra rivers encompassing a length
of 246 km and an average width of 33 km. The total area has been divided into
127 × 893 cells curvilinear grids to perform the 2D hydrodynamic simulation. For
the bathymetry data, the interpolated data from the measured bathymetry of BWDB
for the year 1998 was considered. For the topographic data, Shuttle Radar Topog-
raphy Mission data of the year 2014 was used. Figure 2 shows the model grid and
bathymetry. The boundary data of discharge and the water level have been collected
from BWDB. The observed/generated discharge data of Brahmaputra-Jamuna,
Teesta, Dharla and Dudhkumar River are considered as the upstream boundary condi-
tion. At the same time, observed/generated water levels of Brahmaputra-Jamuna
Fig. 2 Model grid and bathymetry

(Aricha), Old Brahmaputra (Jamalpur), are considered as the downstream boundary

condition.
For this research, two types of conditions have been considered which are base
and hypothetical conditions as shown in Table 1. Three types of scenarios were set up
for the hypothetical conditions—(i) the discharge value of the Teesta-Dharla River
is Zero (Both Zero); (ii) the discharge value of Dharla is Zero (Dharla Zero) and (iii)
the discharge value of Teesta is Zero (Teesta Zero). All models run at the 100-year
return periods condition reflecting the flood of 1998. The boundary condition of the
base condition, BJa Da Ta is shown in Fig. 3.
From the simulation results, flood information such as the area of inundation, depth
of inundation, duration of inundation and flow velocity has been extracted for three
temporal conditions (pre-flood, post-flood and peak flood). Depth of inundation and
Table 1 Scenario considered in this study

Scenarios Criteria
Base condition, BJa Da Ta Flood of 1998
Hypothetical condition
Both Zero BJa D0 T0 Discharge of Dharla and Teesta is zero
But Brahmaputra-Jamuna has the same discharge of the flood 1998
Dharla Zero, BJa D0 Ta Discharge of Dharla is zero
But Brahmaputra-Jamuna and Teesta have the same discharge of the
flood 1998
Teesta zero BJa Da T0 Discharge of Teesta is zero
But Brahmaputra-Jamuna and Dharla have the same discharge of the
flood 1998
a 15000
Brahmputra-Jamuna
Dharla and Dudh Kumar
Brahmaputra-Jamuna Dudhkumar Dharla Teesta

Discharge of Teesta,
100000
Discharge of
10000
(m3/s)
(m3/s)
50000
5000
0 0
1-Jan 20-Feb 11-Apr 31-May 20-Jul 8-Sep 28-Oct 17-Dec 5-Feb
Date
b 20
Old Brahmaputra Brahmaputra-Jamuna
Water level (m
PWD)
0
1-Jan 20-Feb 11-Apr 31-May 20-Jul 8-Sep 28-Oct 17-Dec 5-Feb
Date
Fig. 3 Boundary conditions of the model for Base condition, BJa Da Ta . a Upstream discharge
boundary. b Downstream water-level boundary
Fig. 4 Flood duration calculation process, where blue indicates water in that cell
flow velocity were directly extracted from model results, and duration was calculated
from the depth of inundation. Calculations were done separately for every individual
cell. If one cell contains water continuously for several days, then it was counted
backwards to find the duration for every day in that cell (Fig. 4). After extracting all
the results, a comparison was made between actual and hypothetical conditions to
determine the influence of tributary rivers.
3 Results
3.1 The Flood Characteristics
Before discussing the hypothetical condition results, it is important to examine the

character of the actual flooding (base) condition. Figure 5 shows the flood-hydrograph
characteristics of the considered rivers at the base or actual condition BJa Da Ta . It
is evident from this figure that the Brahmaputra-Jamuna carried more than 90%
(annual average) of flow during the considered year (Fig. 5a). However, during the
peak monsoon, the sharing of Dharla (6.25%) and Teesta (5.26%) increased slightly
(around 1% to 2%). But this slight increase had an impact on local water levels.
Figure 5b shows the 7-day moving average of the rate of rising of water-level hydro-
graph of the Bahadurabad at Brahmaputra-Jamuna, Teesta at Kaunia and Dharla at
Taluk-Simulbari. It is evident from this figure that the rise of Dharla and Teesta
caused an increase of the hydrograph of the Brahmaputra-Jamuna each time. This
phenomenon contributes to the local flooding.
Figure 6 shows the flooding depth and extent for different conditions at peak
monsoon. It confirms that when the river discharge is nil either in Dharla (BJa D0 Ta ),
or Teesta (BJa Da T0 ) or both (BJa D0 T0 , c), the flooding depth is 0 m to 0.30 m lower
than the base condition. Though the areal extent was almost similar in all conditions
(Kurigram 55% and Jamalpur 43%).
a 100%
Percentage of flow
sharing
90%
Brahmaputra-Jamuna Dharla Teesta

80%
1-Jan 1-Feb 1-Mar 1-Apr 1-May 1-Jun 1-Jul 1-Aug 1-Sep 1-Oct 1-Nov 1-Dec
Date
0.4
b
hydrograph (m/day)
0.2
Rate of rise of
-0.2
10
Brahmaputra-Jamuna Dharla Teesta
-0.4
1-Jan 1-Feb 4-Mar 4-Apr 5-May 5-Jun 6-Jul 6-Aug 6-Sep 7-Oct 7-Nov 8-Dec 8-Jan
Date
Fig. 5 The flood-hydrograph characteristics of the considered rivers at the base condition BJa Da Ta .
a. The percentage of flow sharing in actual condition, b. 7-days moving average of rate of rising of
water-level hydrograph of the considered rivers
Fig. 6 Water depth (in m) of a base BJa Da Ta , b both zero BJa D0 T0 , c Dharla zero BJa D0 Ta , d
Teesta zero BJa Da T0 condition at peak-flooding time
3.2 Flood Depth
The simulation results showed that at base condition, BJa Da Ta , the average flooding
depth varies between 2.06 m and 3.37 m in the study domain as shown in Table
2. However, at Both Zero condition, the depth was lowered by 3.8% and 3.6% in
Table 2 Flooding depth of

Scenarios Average flooding depth (in m)
the considered cases the study
domain Pre-monsoon Monsoon Post-monsoon
Base condition, 2.06 3.37 2.15
BJa Da Ta
Hypothetical condition
Both zero 1.98 3.25 2.11
BJa D0 T0
Dharla zero, 2.03 3.30 2.14
BJa D0 Ta
Teesta zero 2.02 3.32 2.12
BJa Da T0
pre-monsoon and monsoon periods correspondingly. In the post-monsoon season, the

depth was similar to that of the base condition (1.86%). The impact of flooding depth
in BJa D0 Ta and BJa Da T0 conditions was prominent in monsoon time. In Dharla
Zero condition BJa D0 Ta , the depth was decreased by 2.07% whereas in Teesta Zero
condition BJa Da T0 , it was found to be 1.48% less than that of the base condition.
The special comparison of flood depth among the base, BJa Da Ta and hypothetical
scenarios (Both Zero BJa D0 T0 , Dharla Zero BJa D0 Ta , Teesta zero BJa Da T0 ) is
shown in Fig. 7. It indicates that the floodwater depth increases 0 m to 0.3 m due to
both Dharla and Teesta River discharge. Whereas, the contribution to Jamuna flood
depth for only Dharla or Teesta is 0 to 0.15 m (Fig. 7). The maximum variation found
in peak flood conditions and the variation in post-flood conditions are negligible.
Here, the area-specific analysis was done by differentiating the flooding of inland
and char areas in the two districts. At base condition, the average depth of flooding
during the monsoon season was 5.44 m and 0.30 m in Kurigram district’s char
land and inland area, respectively. In pre-monsoon and post-monsoon seasons, the
flooding depth was 67% and 59% lower in char areas, whereas the depth was 0.07 m in
both pre and post-monsoon seasons for inland areas. In Jamalpur, the flooding depth
was higher in inland areas (0.34 m) but lower in char areas (5.22 m) in comparison
to the Kurigram district during monsoon season. In the char areas of Jamalpur, the
depth was 70% and 57% lower than the monsoon season in pre and post-monsoon
seasons, respectively. In inland areas of Jamalpur, the depth was 0.10 m in both
pre and post-monsoon periods. Figure 8 shows the flood depth difference between
the base condition and all hypothetical conditions. It is evident from this figure that
the alteration of tributary discharge affects the flooding depth of char areas in both
districts mostly. The maximum difference (0.22 m) was found in the chars of Jamalpur
district in Both Zero BJa D0 T0 conditions during monsoon time. In the case of the
chars of Kurigram, it was 0.12 m. The influence of Teesta was greater (0.07 m in char
and 0.1 m in inland) in flooding depth than Dharla compared to the base condition.
240
Influence of Teesta Influence of Dharla Combined influence of Dharla and Teesta

(BJa Da Ta − BJa D T0 ) (BJa Da Ta − BJa D0 T ) (BJa Da Ta − BJa D0 T0 )
a
Pre-monsoon
e
h
b
Monsoon
f
c
Fig. 7 Difference of flood depth between base and hypothetical condition

Post-monsoon
M. M. Hussain et al.
Combined Influence Influence Combined Influence Influence Combined Influence Influence

influence of Dharla of Teesta influence of Dharla of Teesta influence of Dharla of Teesta
of Dharla (BJa Da Ta − ( − of Dharla (BJa Da Ta − (BJa Da Ta − of Dharla (BJa Da Ta − (BJa Da Ta −
and Teesta BJa D0 T ) 0) and Teesta BJa D0 T ) BJa D T0 ) and Teesta BJa D0 T ) BJa D T0 )
(BJa Da Ta − (BJa Da Ta − (BJa Da Ta −
BJa D0 T0 ) BJa D0 T0 ) BJa D0 T0 )
Pre-monsoon Monsoon Post-monsoon
Fig. 8 Flood depth difference between base and hypothetical condition
3.3 Flood Duration
Duration is one of the major hazard parameters for floods. The study area consists of
inland and char areas. The flood duration was found to be relatively low for inland
areas, whereas the chars experienced very high duration flood. The results show that
in the base condition, the flood duration varies from 0 to 115 days. But in Both Zero
condition BJa D0 T0 , the duration was only 1% less. The influence of Dharla (0 days)
was greater than Teesta (1 day) on flood duration. Figure 9 shows the difference
map of flood duration among the considered scenarios in pre-monsoon, monsoon
and post-monsoon periods. It is evident from this figure that the influence of the
tributary discharge was similar during these three timelines. When considering the
discharge of both tributaries to be nil (BJa Da Ta − BJa D0 T0 ), the flooding duration
reduced to two days compared to the base condition (Fig. 9a–c). The effect of the
Dharla on flood duration was insignificant (nearly 1 day) due to its flashy nature of
the hydrograph. But it resulted in local flood in the Kurigram district (Fig. 9d–f).
Due to the influence of Teesta, the flood duration may increase by nearly three days
(Fig. 9g–i). Figure 10 shows the difference in flood duration in char and inland areas
in both districts. This figure depicts that due to the influence of tributary discharge,
the highest flood duration altered around 4 days in char areas and 1 day in inland
areas. The chars of Jamalpur areas are more susceptible to long-duration flooding
compared to Kurigram district’s flood. The influence of tributary discharge inland
flood duration is insignificant (about one day).
242
Influence of Teesta Influence of Dharla Combined influence of Dharla and Teesta

(BJa Da Ta − BJa D T0 ) (BJa Da Ta − BJa D0 T ) (BJa Da Ta − BJa D0 T0 )
a
Pre-monsoon
b
Monsoon
c
Fig. 9 Difference map of flood duration among the considered scenarios

Post-monsoon
M. M. Hussain et al.
Combined Influence Influence Combined Influence Influence Combined Influence Influence

influence of Dharla of Teesta influence of Dharla of Teesta influence of Dharla of Teesta
of Dharla (BJa Da Ta − (BJa Da Ta − of Dharla (BJa Da Ta − (BJa Da Ta − of Dharla (BJa Da Ta − (BJa Da Ta −
and Teesta BJa D0 T ) BJa D T0 ) and Teesta BJa D0 T ) BJa D T0 ) and Teesta BJa D0 T ) BJa D T0 )
(BJa Da Ta − (BJa Da Ta − (BJa Da Ta −
BJa D0 T0 ) BJa D0 T0 ) BJa D0 T0 )
Fig. 10 Flood duration difference between base and hypothetical condition
3.4 Flood Velocity
Figure 11 shows the histogram plot of velocity magnitude in different scenarios.

The x-axis showed the velocity magnitude class while the y-axis shows the number
of grid cells containing that velocity in the model domain. As the study domain
contains the land of various elevations, the range of velocity distribution is quite
large. For example, the near channel area of chars has very low elevation compared
to the floodplain or mid-char areas. Therefore, we consider the histogram plot for
comparing the velocity of different cases. All the subfigures of Fig. 11 show two peaks
due to the variation of velocity in low-lying and higher elevated areas. It depicts that
unlike the flood duration, the flooding velocity shows seasonal variation. In base
condition (BJa Da Ta ), during the pre-monsoon period, the velocity magnitude varies
from 0 m/s to 2.26 m/s. In the inland areas, the velocity varies between 0 and 0.22 m/s,
and when near the channel, it ranges between 0.44 m/s and 0.89 m/s, maintaining
an average velocity of 0.40 m/s. During the monsoon period, the maximum velocity
was found to be 2.51 m/s, wherein the average velocity increases by 1.5 times.
In the post-monsoon season, the velocity distribution was quite similar to the pre-
monsoon season (Fig. 11a–c). The distribution pattern of the hypothetical cases was
similar to that of the base condition (Fig. 11d–l). In the pre-monsoon period, the
velocity magnitude ranges from 0 m/s to 2.3 m/s with the average value of 0.38 m/s
in all hypothetical conditions. In the monsoon season, the maximum velocity was
found to be 2.53 m/s (BJa D0 T0 ) with an average value of 0.60 m/s. In the post-
monsoon season, a maximum velocity of 2.3 m/s (BJa Da T0 ) was noted, wherein
the average velocity was 0.40 m/s. Figure 12 shows the spatial distribution of the
difference of flood velocity magnitude among the considered scenarios. It illustrates
that the contribution in velocity magnitude of Dharla and Teesta varies between −
0.015 m/s and 0.030 m/s. The main variation was observed at the lower left bank of
Brahmaputra-Jamuna in peak-flooding conditions as this section has better floodplain
connectivity (the right bank of the study domain was partially embanked).
Fig. 11 Distribution of velocity magnitude of the considered cases
At base condition, the average velocity was 0.97 m/s in Kurigram district’s char
areas, whereas in Jamalpur it was 0.93 m/s as the valley slope is higher in Kurigram.
But for the inland areas, the flooding velocity was 0.15 m/s in Kurigram and 0.25 m/s
in Jamalpur. As the inland area of Kurigam contains partial embankment, it hampered
the drainage path of the floodwater. The contribution of the tributary in velocity
increment or decrement in char and inland areas was very small (0.01 m/s).
4 Discussion
Brahmaputra-Jamuna River is one of the major rivers in Bangladesh, and a lot of

studies have been conducted on the Brahmaputra-Jamuna River floods (Best et al.
2007; Dewan et al. 2003; Islam and Chowdhury 2002; Mosselman 2006). But no
previous study has been done to investigate the contribution of its tributaries. In this
study, we tried to find the impact of tributaries by comparing different flood param-
eters (inundation depth, flood duration, flow velocity) in different conditions (base,
BJa Da Ta , Both Zero BJa D0 T0 , Dharla Zero BJa D0 Ta , Teesta Zero BJa Da T0 ). The
results show that the inundation depth varies up to 3.8% due to contribution of Teesta
a b c
Combined influence of Dharla and Teesta
(BJa Da Ta − BJa D0 T0 )
d e f
(BJa Da Ta − BJa D0 T )
Influence of Dharla
g h i
(BJa Da Ta − BJa D T0 )
Influence of Teesta
Fig. 12 Spatial distribution of flood velocity magnitude change among the considered scenarios
and Dharla river water. Islam and Chowdhury (2002) claims that, continuous inflow
of high discharge for more than two and half months occurred in the Brahmaputra-
Jamuna in 1998, which is consistent with the base condition BJa Da Ta results. The
contribution in velocity magnitude of Dharla and Teesta varies between -0.015 m/s
and 0.030 m/s. The study also found that Teesta and Dharla only contribute to 10%
of discharge water in Brahmaputra-Jamuna River. The large difference between the
river discharge of Brahmaputra-Jamuna and Teesta and Dharla is the probable reason
for this small impact in Brahmaputra-Jamuna floods.
Although the aerial extent of flooding is largely dependent on the flooding pattern
of the Brahmaputra-Jamuna, tributary flooding affects local flood conditions espe-
cially in the char areas. The flooding depth increase by up to 0.20 m due to the
flooding conditions of the tributaries. The flooding duration can increase by up to
three days due to the influence of the Teesta and Dharla. The flooding velocity was
influenced by the presence of the local infrastructure (e.g. embankments). The down-
stream district Jamalpur was found to be more susceptible to flooding. The recent
flood of the year 2020 has also shown a similar damage trend in Jamalpur (NAWG
2020).
This study aimed to gather new insights into the impact of tributaries on the
Brahmaputra-Jamuna floods, by undertaking analyses that have not been conducted
before. However, this study only accounted for Brahmaputra-Jamuna high flood (100-
year return periods) condition. For a comprehensive understanding of the nature of
Brahmaputra-Jamuna’s tributaries, other types of scenarios (such as the high flood
in Dharla-Teesta, normal flood in Jamuna etc.) can be investigated.
5 Conclusion
In this study, we investigate the influence of the flow of Teesta and Dharla river
tributaries on flooding of the Brahmaputra-Jamuna. The study found that the rise of
water level in the Dharla and Teesta rivers cause an increase in the water level of
the Brahmaputra-Jamuna each time, which results in local floods, especially in char
areas. The impact of Teesta is higher in terms of flood depth, duration and velocity
compared to Dharla River. The downstream reach (i.e. Jamalpur district in this case)
is more susceptible to tributary flooding. The char and inland flooding characteristics
are significantly different. The flooding depth increases almost 25 times in char lands
relative to inland flooding. The flood duration increases nearly 18 times in char lands
and velocity may increase ninefold in char areas compared to inland areas. Therefore,
similar flood adaptation techniques in inland and char areas may be inappropriate.
It is expected that the results of this study will contribute to Brahmaputra-Jamuna’s
long-term flood risk management planning.
Acknowledgements This work was carried out under the Institute of Water and Flood Manage-
ment (IWFM), BUET project titled “Developing Institutional Framework for Flood Preparedness
Program (FPP)” funded by the National Resilience Programme (NRP) of the Government of
Bangladesh with the technical assistance of United Nations Development Programme (UNDP),
UN Women and United Nations.
References
Albano, R., L. Mancusi, A. Sole, and J. Adamowski. 2017. FloodRisk: A Collaborative, Free
and Open-Source Software for Flood Risk Analysis. Geomatics, Natural Hazards and Risk 8:
1812–1832. https://doi.org/10.1080/19475705.2017.1388854.
Archana, S., G. RD, and S. Nayan. 2012. RS-GIS Based Assessment of River Dynamics of Brahma-
putra River in India. Journal of Water Resource and Protection. https://doi.org/10.4236/jwarp.
2012.42008.
Best, J.L., P.J. Ashworth, M.H. Sarker, and J.E. Roden. 2007. The Brahmaputra-Jamuna River,
Bangladesh. Large Rivers: Geomorphology and Management: 395–430.
Bhuiyan, S.R. 2014. Flood Hazard and Vulnerability Assessment in a Riverine Flood Prone Area:
A Case Study.
Dewan, A., M. Nishigaki, and M. Komatsu. 2003. Floods in Bangladesh: A Comparative Hydrolog-
ical Investigation on Two Catastrophic Events. Journal of the Faculty of Environmental Science
and Technology, Okayama University 8: 53–62.
Dewan, T.H. 2015. Societal Impacts and Vulnerability to Floods in Bangladesh and Nepal. Weather
and Climate Extremes 7: 36–42. https://doi.org/10.1016/j.wace.2014.11.001.
Ferdous, M.R., A. Wesselink, L. Brandimarte, G. Di Baldassarre, and M.M. Rahman. 2019. The
Levee Effect Along the Jamuna River in Bangladesh. Water International 44: 496–519. https://
doi.org/10.1080/02508060.2019.1619048.
FFWC. 2018. Annual Flood Report. Bangladesh Water Development Board (BWDB).
Islam, A., and J.U. Chowdhury. 2002. Hydrological Characteristics of the 1998 Flood in Major
Rivers.
Karim, M.F., and N. Mimura. 2008. Impacts of Climate Change and Sea-Level Rise on Cyclonic
Storm Surge Floods in Bangladesh. Global Environmental Change, Globalisation and Environ-
mental Governance: Is Another World Possible? 18: 490–500. https://doi.org/10.1016/j.gloenv
cha.2008.05.002.
Klaassen, G., and K. Vermeer. 1988. Confluence Scour in Large Braided Rivers with Fine Bed
Material. In Proceedings of the International Conference on Fluvial Hydraulics, pp. 395–408.
Budapest, Hung: Int. Assoc. Hydraul. Res.
Kvočka, D., R.A. Falconer, and M. Bray. 2016. Flood Hazard Assessment for Extreme Flood Events.
Natural Hazards 84: 1569–1599. https://doi.org/10.1007/s11069-016-2501-z.
Lesser, G.R., J.A. Roelvink, J.A.T.M. van Kester, and G.S. Stelling. 2004. Development and Vali-
dation of a Three-Dimensional Morphological Model. Coastal Engineering 51: 883–915. https://
doi.org/10.1016/j.coastaleng.2004.07.014.
Mamun, M.A., N. Huq, Z.F. Papia, S. Tasfina, and D. Gozal. 2019. Prevalence of Depression Among
Bangladeshi Village Women Subsequent to a Natural Disaster: A Pilot Study. Psychiatry Research
276: 124–128.
Mechler, R., and L.M. Bouwer. 2015. Understanding Trends and Projections of Disaster Losses
and Climate Change: Is Vulnerability the Missing Link? Climatic Change 133: 23–35. https://
doi.org/10.1007/s10584-014-1141-0.
Mosselman, E. 2006. Bank Protection and River Training Along the Braided Brahmaputra-Jamuna
River, Bangladesh. Braided Rivers: Process, Deposits, Ecology and Management 36: 279–287.
Munich Re. 2015. Weather Catastrophes and Climate Change. Münchener Rückversicherungs-
Gesellschaft.
Nasreen, M. 2004. Disaster Research: Exploring Sociological Approach to Disaster in Bangladesh.
Bangladesh e-Journal of Sociology 1: 1–8.
NAWG. 2020. Monsoon Floods 2020, Coordinated Preliminary Impact and Needs Assessment.
Needs Assessment Working Group (NAWG), Bangladesh.
Pal, P.K., A. Rahman, and A. Yunus. 2017. Analysis on River Bank Erosion-Accretion and Bar
Dynamics Using Multi-Temporal Satellite Images. American Journal of Water Resources 5:
132–141.
Paul, B.K. 2011. Environmental Hazards and Disasters: Contexts, Perspectives and Management.
John Wiley & Sons.
Pramanik, M. 1991. Environmental Vulnerability in Bangladesh-Paper Presented in the UNDP. In
ESCAP Workshop on Regional Development Issues and Environment, 18–21. Dhaka: Bangladesh,
November.
Rahman, M.M., and M.M. Ali. 2016. Flood Inundation Mapping of Floodplain of the Jamuna River
Using HEC-RAS and HEC-GeoRAS. Journal of PU 3: 24–32.
Schumm, S.A., and B.R. Winkley. 1994. The Variability of Large Alluvial Rivers. ASCE.
Shampa, 2019a. Hydro-Morphological Study of Braided River with Permeable Bank Protection
Structure. Doctoral Thesis, Kyoto University 102.
Shampa, 2019b. Hydro-Morphological Study of Braided River with Permeable Bank Protection
Structure. Thesis, Kyoto University, Japan 153.
Shampa, Y. Hasegawa, H. Nakagawa, H. Takebayashi, and K. Kawaike. 2017. Dynamics of Sand
Bars in Braided River: A Case Study of Brahmaputra-Jamuna River. Journal of Japan Society of
Natural Disaster Science 36: 121–135.
Uddin, K., B. Shrestha, and M.S. Alam. 2011. Assessment of Morphological Changes and Vulner-
ability of River Bank Erosion Alongside the River Jamuna Using Remote Sensing. Journal of
Earth Science and Engineering 1.
Younus, M.A., R. Bedford, and M. Morad. 2007. Food Security Implications of Failure of
Autonomous Crop Adaptation to Extreme Flood Events: A Case Study in Bangladesh.
Water Infrastructure and Development
Hybrid Coast Protection Approach
in Bangladesh: A Case Study
on Effectiveness of Small-Scale Forest
in Reducing Surge Induced Inundation
and Supporting Local Livelihoods
Mita Kazi Samsunnahar
Abstract Polders are subject to significant threats and often fail to protect the hinter-
land during cyclone induced storm surges. This study investigated the impact of
mangrove afforestation in attenuating storm surge impacts on the embankment, as
well as the social benefits of adapting this measure. Hydrodynamic model Delft-
3D and ArcGIS 10.3 have been used to simulate, calculate and visualize coastal
flood inundation condition. Primary and secondary source information were used
to validate the model and assess social benefits of afforestation. For polder 47/4,
the presence of 1.61 km2 afforestation on the foreshore, protected additional 3 km2
land from inundation during cyclone Sidr. While increasing polder height to 7 m
slightly influenced inundation extent by reducing it 0.3 km2 , additional afforestation
of the same roughness as exists now, helped to further reduce inundation depth in
the hinterland. Social benefits provided by the existing forest were also considerable
as local people collect fuelwood and Nipa palm regularly. While fuelwood was used
for non-commercial uses, Nipa palm, on the other hand, was used by poor village
women for preparing mats to sell at the local market. From the collected data, net
benefit of this forest was calculated to be Tk. 5990.68 ($70.20)/hectare/year.
Keywords Storm surge · Coastal protection · Nature-based infrastructure ·

Afforestation · Social benefits
1 Introduction
As a vast majority of the world population lives near the coast (Nichols and Small
2002), protecting coastal zones from adverse climatic phenomena is a major concern.
Over the last few decades, climate change risks, extreme weather conditions, coastal
flooding etc. have been increasing and they are expected to further worsen in the
future. Traditionally, coastal areas of the world have been protected by engineering
structures which exhibit high effectiveness initially. But gradually, interference with
M. K. Samsunnahar (B)
Davidson Laboratory, Stevens Institute of Technology, Hoboken, USA
252 M. K. Samsunnahar
surrounding nature (De Vriend et al. 2015) and incapability to keep up with rapidly
changing coastal conditions (Sutton-Grier et al. 2015), often undermines their orig-
inal purpose (Jones et al. 2012). This has made researchers look for more sustain-
able and effective ecosystem-based solutions to the issue. Ecosystem-based Adapta-
tion entails a wide range of ecosystem management activities targeted at increasing
resilience and reducing vulnerability of people and the environment. Many new
nature-based concepts with different names (De Vriend et al. 2015; Beck and Shepard
2012; Cheong et al. 2013; NYS 2013; Ferrario et al. 2014; Arkema et al. 2013) have
been introduced, as recent global catastrophes, such as the Indian super cyclone, and
hurricanes Katrina, Irene and Sandy, have raised awareness regarding the challenges
and potential mitigation plans to protect people and property (Das and Vincent 2009;
Reddy et al. 2016). Pure nature-based approaches are often not feasible because they
take substantial amount of time and space to implement (Sutton-Grier et al. 2015).
Using a combination of both built and natural infrastructures, it provides a scope to
capitalize on the strengths of both, minimizing the weaknesses of each. This type of
hybrid approach incorporates the intentioned use of natural and engineered features
to produce engineering functions in combination with ecosystem services and social
benefits (Bridges et al. 2013; Simm et al. 2016; Fink 2016), where “bio-based” solu-
tions alone are likely to be ineffective. Examples from the USA (Heerden and Ll
2007) and the UK (van Slobbe et al. 2013) have shown tremendous performance of
hybrid infrastructures during cyclones and regular wave action. Since the concept of
hybrid infrastructure is new, no research has yet been conducted in the context of
Bangladesh on this topic.
Bangladesh is widely recognized as one of the most disaster-prone countries of
the world with its repetitive cycle of floods, cyclones and storm surges. The coastal
zone is particularly vulnerable to climate-related impacts, especially cyclones, with
a severe cyclone striking the Bangladesh coast every three years (GoB 2008). Loss of
human lives and destruction of assets during these cyclones are primarily attributed
to surge waves (Chowdhury et al. 1993). Major cyclonic storms of 1970 and 1991
together caused death of nearly half a million people and economic losses exceeded
US$200 billion (Quadir 2008). Polders have always been under significant threat
during cyclone, and failure of polders has caused severe damages to the housing
and agricultural sector. During cyclone Sidr in 2007, total damages to over 2000 km
of embankments and other critical water control structures amounted to about BDT
4.9 billion (US$71 million) (GoB 2008). The total length of embankment of 49
sea facing polders is 957 km, of which only 60 km has forest belts; and in many
areas, the existing forest belt is degraded (Dasgupta et al. 2010). So, it is essential
to take measures for enhancing the degree of protection of polders by providing
afforestation.
Under the Coastal Embankment Rehabilitation Project (CERP- II), it was seen
that afforestation in the foreshore area can play a very important role in the protec-
tion of embankments against wave action. Placed between “hard measures” like
embankments, and the open sea greenbelts can help in reducing the impact of storm
surge on the embankments, thereby increasing their lifespan and reducing the cost of
maintenance and augmentation (CEGIS 2016). Moreover, it also has the potential of
Hybrid Coast Protection Approach in Bangladesh… 253
providing significant amount of revenue through harvesting of forest resources. So,

in the Coastal Embankment Improvement Project (CEIP-I), along with increasing
polder height, Bangladesh Water Development Board (BWDB) has planned limited
afforestation on the foreshore to protect the embankment and reduce maintenance
cost.
Coastal greenbelts have long been seen as an important strategy for reducing the
vulnerability of coastal populations to climate-related hazards in Bangladesh, and the
country has over five decades of experience in coastal afforestation and reforestation.
Over 140,000 ha of mangroves have been planted as social forestry along the coast
since the 1960s (UNDP 2015). Also, to combat the adverse effects of climate change
induced coastal hazards, the Bangladesh Forest Department (BFD) has initiated the
afforestation project titled Climate Resilient Participatory Afforestation and Refor-
estation Project (CRPARP). While BWDB and BFD are separately taking measures
to protect the coastal region, their efforts to integrate coastal embankment and coastal
greenbelt planning and management have been limited till date (UNDP 2015). This
study aims to demonstrate the possible outcomes of such an integration.
Also in delineating afforestation, previous studies have only considered width
of afforestation, orientation of greenbelt and suitability of trees (CEGIS 2016) as
design criteria for coastal belt. But if resistance provided by designed afforestation
was considered, it would provide a smore comprehensive view about attaining desired
protection during cyclonic storm surges. This study aims to take into consideration
this important parameter along with afforestation width, to assess the effectiveness of
hybrid intervention (integration of coastal embankment and foreshore afforestation)
in providing protection.
Although several studies have been conducted to quantify the benefits of Sundar-
bans mangrove forest (Abdullah et al. 2016; Islam et al. 2019; Slam et al. 2020),
very few studies have focused on community-based afforestation in the coastal belt
of Bangladesh (Chow 2015). Understanding people’s attitudes towards this type
of small-scale afforestation project may help design them as part of social forestry
programmes. Inclusion of benefits assessment of local people will increase the accept-
ability of hybrid intervention as a long-term adaptive measure to protect coastal
communities from cyclonic hazards in Bangladesh.
2 Study area
This study was conducted in polder 47/4, situated at Kalapara upazila in Patu-
akhali district, which is a highly disaster-prone area of Bangladesh. Kalapara upazila
(located at 21.9861°N and 90.2422°E) has a total area of 492.102 km2 , bounded by
Amtoli upazila on the north and west, Rabnabad channel on the east and the Bay
of Bengal to its south. The polders in Kalapara upazila are presented in Fig. 1. The
Polder 47/4 is bounded by polder 46 on the north, polder 47/1, 47/2 and 47/3 on the
west, polder 47/5 and Rabnabad channel on the east and Kuakata and the Bay of
Bengal on the south (Fig. 1).
Fig. 1 Map of study area
Polder 47/4 was selected for this study due to its exposure to cyclones and cyclone
induced storm surges. Because of its proximity to the sea and poor embankment
condition, local people are severely affected by cyclones, storm surges and salinity
intrusion, making this area as one of most climate vulnerable areas of Patuakhali
(GoB 2013). In a reconnaissance survey, local people confirmed breaching of polder
47/4 at different points during Cyclone Sidr. Cyclones of moderate intensity like Aila
and Mohasen have also damaged resources of this area. Since 1966, this polder has
been performing as a defence mechanism against tidal flooding and storm surges. But
in all these years, apart from occasional minor maintenance work, no noticeable reha-
bilitation work has been done. Thus, this polder has become vulnerable to extreme
forces like storm surges. For this study, polder 47/4 represents a highly vulnerable
exposed polder from the coastal region of Bangladesh. This polder has a small-
scale mangrove forest, locally known as Dhulaswar Forest, consisting of mature
trees along the foreshore of the embankment, which provides an ideal condition for
conducting this research. The outcomes of this research will help to understand the
general role of mangrove as an additional protective measure during extreme cyclonic
events. Inclusion of benefits assessment of forest goods, along with assessing the role
of mangrove plantation in reducing storm surge induced flood inundation for polder
47/4, will depict the suitability of hybrid intervention as a long-term adaptive measure
at the community level.
3 Methodology
This study employs a technical perspective to investigate the effectiveness of

mangrove plantation in the foreshore of the earthen embankment in protecting coastal
communities from storm surge impacts. It also focuses on the social benefits of a
small afforestation project in polder 47/4 by quantifying the services this small man-
made forest provides to the local community. In this section, the methodology of
data collection, description of models and tools and techniques related to the study
are described.
3.1 Model Description
In this study, a calibrated and validated model of Bangladesh coast is set up using
the Delft-3D hydrodynamic model, and then used to study the impact of mangrove
afforestation in attenuating the devastating impact of storm surges. Delft3D hydrody-
namic model is a computational tool which can simulate flow, sediment transports,
waves, water quality, morphological developments and ecology for coastal, river
and estuarine areas. Delft3D-FLOW solves the unsteady shallow water equation in
two dimensions (depth-averaged) or in three dimensions (Delft Hydraulics 2006).
Delft3D flow model coupled with the cyclone generated from the Delft-Dashboard is
utilized to simulate coastal flooding for different cyclones. The data extracted from
simulation has been processed using ArcGIS to produce maps.
3.2 Basic Model Setup
Delft-3D model domain extends from 21°03 ‘00 “N to 24° 00 ‘00 “N and 86°00
‘00 “E to 94°00 “00 ‘E. The open sea is bounded parallel to latitude 20°00 ‘00 “N,
and the land domain considers all the 19 coastal districts of Bangladesh. A detailed
description of model domain, upstream and downstream boundary conditions, land
topography, river bathymetry, land use and land cover data can be found in the study
conducted by Al Azad et al. (2018). The model has been validated for Cyclone Mora
(Al Azad et al. 2018), and cyclone Sidr (Mita 2019).
3.3 Modification of the model
Since the study area is a segment of the entire model, to modify this portion according
to field data, several parameters such as polder height, width, actual polder height,
mangrove afforestation properties (length, width, composition and density) etc. are
collected from the field for incorporating in the coastal model. In addition to field
measurements, several Focus Group Discussions (FGDs) and Key Informant Inter-
views (KII) have been conducted. As Delft-3D considers impact of land cover as
resistance, Manning’s roughness coefficient is calculated from collected properties
of trees. Detailed description of these methods will be discussed in this section.
3.3.1 Polder Height Modification
To better represent the existing height of polder in the model, primary data related to
polder characteristics (width, slope etc.) is collected from several locations of polder
47/4. From collected data, height of polder is modified in the coastal model which
initially had design height as input.
3.3.2 Quantification of Mangrove Plantation Properties
As for afforestation, general ideas regarding the area of the plantation, vegetation
composition, vegetation characteristics etc. were acquired from Key Informant Inter-
views (KII) with foresters. Vegetation characteristics and vegetation composition
were then observed during the transect walk. Later, using Google map/Satellite
image, area of afforestation is determined again along with its length and width.
The density of the plantation is represented as Manning’s roughness coefficient in
DELFT-3D model. In this study, Manning’s roughness coefficient of existing forest is
determined by using a direct technique (Chow 1959). In a densely wooded floodplain,
size of tree trunks is a major contributing factor in determining roughness coefficient.
In this type of locations, vegetation density can be calculated by counting the number
of trees and measuring the trunk size of the trees in a representative sample area. The
n value can be computed by using Eq. 1.
2
c∗ Ai 1
n = n0 1 + R 4/3 (1)
2g AL n0
Where,
n 0 = Manning’s boundary-roughness coefficient excluding the effect of the
vegetation (a base n),
C∗ = The effective-drag coefficient for the vegetation in the direction of the flow,
Ai =The total frontal area of vegetation blocking the flow in the reach,
G = The gravitational constant,
A = The cross-sectional area of the flow,
L = Length of the channel reach being considered, and.
R = The hydraulic radius.
Cowan (Chow 1959) developed a procedure for estimating the effects of these
factors to determine the value of n for a channel. By altering Cowan’s procedure, the
following equation is developed to compute the value of n for floodplain (Arcement &
Schneider 1989),
n0 = (nb + n1 + n2 + n3 + n4 )m (2)
where,
nb = a base value of n for floodplain’s natural materials, n1 = a correction factor
for the effect of surface irregularities on the floodplain, n2 = a value for variations in
shape and size of the floodplain cross section, (value = 0 for floodplain), n3 = a value
for obstructions on the floodplain, n4 = a value for vegetation and flow conditions
on the floodplain and m = a correction factor for meandering of the channel (Value
= 1 for floodplain).
Vegetation density calculation
Another required parameter for calculating Manning’s n, vegetation density ( ALAi
term in Eq. 1), is determined for each subsection. To determine the roughness coef-
ficient more accurately, data is collected from two subsections of the forest. Subsec-
tions are chosen based on spatial variance in tree density. A representative sample
area of 30 m by 15 m is chosen to represent the roughness of the subsections accu-
rately. The floodplain is divided into subsections based on spatial geometric and (or)
roughness differences in different parts of the forest. In this study, two sample areas
of 15 m by 15 m along the cross section and in the flow direction are taken to calculate
roughness. The representative sections are selected based on the recommendation of
local foresters for obtaining close to accurate representative vegetation density. A
section of 15 m length perpendicular to the flow direction is marked in the sample
area. Every tree within 7.5 m along either side of this marked line is counted. The
total area occupied by the trees ( Ai ) in the sampling area can be computed from
the number of trees,their diameter and the depth of flow in the floodplain. Once
the vegetation area, Ai , is determined, the vegetation density can be computed by
using following equation,

Ai h n i di
V egg = = (3)
AL hwl
where,

n i di = the summation of number of trees multiplied by tree diameters, h =
height of water in floodplain, w = width of sample area and l = length of sample
area.
3.4 Assessing Benefits of Livelihood Groups
To ensure sustainability of the plantation, it is very important to involve the local

community in the process of operation and maintenance. That will only be possible
if the plantation can be considered as community resources. Quantification of
extractable benefits from small-scale plantation can help policy makers, relevant
implementing agencies and the local people to better understand its impact the on
local economy, and accordingly design the plantation in a manner whereby locals
can get benefited without the main purpose of the plantation being harmed.
To assess the benefits of different local people from mangrove plantation, Partic-
ipatory Rural Appraisal (PRA) tools, such as, Focus Group Discussions (FGD), Key
Informant Interviews (KII) have been used. Field visits were made to the study area
to collect data about the dependency of people on forest resources. In three different
Focus Group Discussions with local people living in proximity to forest areas, data
regarding the quantity of different goods extracted from forest, purpose of collec-
tion and the market value of each good were collected. Total collection quantities,
costs, gross values and net values of fuelwood and other goods are calculated from
estimated harvests. Seasonal variation in the extraction of materials from the forest
has also been given emphasis while collecting data. Also, during FGDs, variation in
dependency is also sorted into three different categories. During several Key Infor-
mant Interviews with residents as well as local foresters, data collected during FGD
are cross-checked. Later using all collected data, the formulation of total net value
density, or periodic net benefits (PNB) per hectare, for a forest patch a is calculated
using the conceptual model developed by Chow (2015). The fundamental equation
is:

PNBa ha= (Pa − C G (Da ))X g − Fg a,i − Ma Ha (4)
i g
Where for every good g and collector i,

Pg = the market price, CG = the variable production cost (e.g. transportation cost),
Xg = the extracted quantity, Fg = the fixed production cost, Ma = the forest manage-
ment cost (in this case paid by government),Ha = the patch size and Da = the distance
of forest patch from the village.
4 Results and discussions
The results of this study are presented in two parts: the first part will focus on
the numerical simulations of several scenarios to demonstrate how the presence of
plantation can influence storm surge induced flooding during Cyclone Sidr (IMD
2021), which is one of the most devastating cyclones in the history of Bangladesh
(JSCE 2008a) and the most destructive one for polder 47/4. In the second part,
the benefits of residents from this small-scale plantation will be presented to address
several important yet understated factors, that are important for designing this hybrid
approach in a sustainable manner.
4.1 Effectiveness of Plantation on Reducing Surge Induced

Flooding
The roughness value for the floodplain is determined by selecting a base value (nb )
for the natural bare soil surface of the floodplain, and then adding adjustment factors
due to vegetation density, irregularity in surface and obstructions in flow path by
using Eq. 2. The adjustments are made based on observation, judgement, experience
and guidance from previous studies (Chow 1959). For this study, the base roughness
(nb ) is assumed to be 0.025 which is consistent with assumptions made by Jisan
et al. (2018) and Zhang et al. (2012). Using Eq. 2, Manning’s boundary-roughness
coefficient excluding the vegetation in two subsections are calculated to be 0.31
and 0.53, respectively, after making reasonable assumptions. The values of vege-
tation density from field measurements are 0.35 and 0.5, respectively. Then using
Eq. 1, the Manning’s roughness coefficient values for two subsections are 0.11 and
0.13. For simplicity, the average of two values 0.12 has been used in the model to
represent the forest. The polder height has been measured in several locations and the
average measured value of 4.51 m has been incorporated in the model. As mentioned
earlier, Cyclone Sidr making landfall to the west of the study area is simulated to
assess the effectiveness of plantation in reducing storm surge impacts. Generally,
the portion of coastline which experiences onshore winds, usually experience high
surge levels and vice versa (Needham and Keim 2011). So, ideally, if the forest is
present in the onshore wind direction, it will provide additional resistance to the
storm surge. Cyclone Sidr brought onshore winds and a high storm surge on the
existing Dhulaswar forest, and thus it is considered as an ideal scenario for studying
the effectiveness of afforestation in reducing storm surge impacts. In case of cyclones
making landfall in the central and eastern region of the Bangladesh coast, the wind
will blow in the offshore direction during cyclone landfall near Dhulaswar forest.
This is due to anticlockwise rotation of the cyclone system. So theoretically, the area
will experience negative to no surge. But due to weather anomalies, the water level
may get higher than usual in few post-landfall tide cycles, and in this case, the forest
will provide additional resistance. The forest also protects the embankment from
regular tidal water-level fluctuations and wave actions. Assessing the effectiveness
of afforestation during tidal conditions and various wave actions is beyond the scope
of this study.
4.2 Scenario One: Role of Existing Afforestation
Inundation map from the model generated data is presented below (Fig. 2) for this
scenario. From this map, it is observed that heavy inundation has occurred in the
north-eastern side of the afforestation. Maximum inundation depth is more than 3 m.
Slight inundation (depth 0.01–0.5 m) is observed in the land behind the afforestation.
Distribution of inundated land under determined categories is given in Fig. 7. Here,
Fig. 2 Inundation map of polder 47/4 considering existing afforestation
inundation free land inside the polder is about 63.69 km2 . About 6.63% of the land
inside this polder is inundated during cyclone Sidr.
During the field visits, local people verified that the flooding occurred from the
exact location where the model simulated result shows significant inundation during
cyclone Sidr (Fig. 3). As a result, the polder was breached, and later they repaired
the polder. After a year of Sidr, while doing maintenance work, BWDB also took
additional measures to protect the slope. When asked about water depth during Sidr,
responses from locals varied widely. Some claimed that the highest water depth
reached about 10–12 ft (3.05–3.66 m), while others suggested the highest level to be
15–18 ft (4.57–5.49 m). However, the polder started breaching when the water level
reached about 4.5 ft. (1.37 m), according to people residing near the polder.
4.3 Scenario Two: Without Afforestation on the Foreshore
In this hypothetical scenario, the role of Dhulaswar mangrove forest is tested by

assuming absence of plantation on the foreshore of the polder. It can be observed from
Fig. 4 that the inundated area has increased from the previous scenario. Especially
behind the imaginary barren land, the inundation has increased, which implies that
presence of afforestation had previously reduced inundation in this portion. Overall,
Fig. 3 Inundated area from model and image taken from field visit of same location
10.60% of the area under polder 47/4 is inundated in the absence of afforestation on
the foreshore. The total inundation free area has reduced to about 60 km2 (Fig. 7).
Comparing Figs. 3 and 4, it can be said that the presence of small-scale mangrove
forest on the foreshore has played a significant role in reducing inundation in polder
47/4. Presence of this small-scale forest saved additional 3 km2 land from surge
induced inundation.
4.4 Scenario Three: Conventional Coast Protection

Approach (Actual Afforestation with Polder Height 7 m)
Traditionally, the Bangladesh coast has been protected by polders from storm surges
and tidal flooding. Under the CEIP project of BWDB, the adjacent polder (polder
48) is to be raised as high as 7 m to protect the area inside the polder from coastal
flooding, as the current height has proved to be inadequate in past few events. Since
neither the present polder height nor the afforestation is deemed adequate to provide
protection from storm surges during cyclone Sidr, this scenario assesses the effect
Fig. 4 Inundation map of polder 47/4 without afforestation on the foreshore
an increase in polder height can have on flooding conditions. So, the polder is taken
as 7 m high for this scenario and pictorial representation of the results are given in
Fig. 5.
Here, as it is seen in Figs. 2 and 5, the rise of polder from 4.73 m to 7 m does not
exhibit a substantial change in inundation free area. When polder height is increased,
the inundation free area has increased to slightly higher than the existing polder
with the afforestation on the foreshore (63.39 to 63.71 km2 ), reducing area of low
inundation water depth (Fig. 7). Area under high water depth remains unaffected
although the polder height is increased by almost 2.5 m.
4.5 Scenario Four: Hybrid Intervention (Increased

Afforestation Area with Existing Polder Height)
As the polder height of 7 m is not adequate to prevent flooding inside the polder,
in this case, the most affected portion on polder is covered by about 1.2 km width
of hypothetical plantation on the foreshore, keeping the polder height 4.5 m. The
width is determined considering width of existing forest and available land on the
foreshore. Here, the map shows that the afforestation is very effective in decreasing
inundation depth in the vulnerable portion of the polder (Fig. 6). The maximum
water depth plummeted to below 1 m from previous depth of about 3 m inside the
Fig. 5 Inundation map of polder 47/4 increasing polder height to 7 m along with existing
afforestation on the foreshore
Fig. 6 Inundation map of polder 47/4 increasing afforestation area on the foreshore along with
existing polder having height of 4.73 m
100.00
Barren land
Actual condi on
10.00
Polder height 7m
Inundated area (Km2)
Increased afforesta on
1.00
0.10
0.01
InundaƟon Depth (m)
Fig. 7 Classification of inundated area for different inundation water categories
polder. From Fig. 6, it is found that, increased afforestation has significant impact in
reducing higher surge depth inside the polder.
4.6 Comparison of Scenarios
For polder 47/4, it is true that existing afforestation exhibits a positive impact on
attenuating storm surge inundation inside the polder during cyclone Sidr, as inun-
dation free area increased from 60.70 to 63.40 km2 with the presence of existing
afforestation. In other words, if Dhulaswar reserved forest was not present in the fore-
shore, the inundated area would be about 3 km2 area more. Conventional approach
of increasing polder height, however, does not play a significant role in reducing
inundation extent for this polder for cyclone Sidr. A polder height of 7 m can free
63.7 km2 areas from inundation, which is 0.3 km2 more than the inundation free
land for existing condition. So, only increasing polder height may not be the most
viable solution to keep the polder area protected. While there is a change in overall
inundation area, as the modelled forest is same as found in the field, efficacy of
presence of plantation is not properly visible, as the most vulnerable portion of the
polder is beyond forest coverage. Because of that, in the following Fig. 6, for first
three scenarios, variations in inundated area for higher water depths (1.5 m–6.0 m)
have not changed. When the vulnerable portion of polder is covered by forest, area
under higher water depth (1.5 m–6.0 m) reduced (Fig. 7) significantly, although a
noticeable reduction in inundation free area is not observed. Local complex hydro-
dynamics and intricate network of water bodies often play a pivotal role in deter-
mining the efficacy of afforestation measures. Even if the spatial extent (horizontal
extent) of flooding remains unaffected, water depth over land (vertical extent) can be
significantly reduced by afforestation. In this case, proposed increased afforestation
scenario is very effective.
4.7 Social Benefits of Mangrove Afforestation
Afforestation does not only function as a shield during storm surges, it also provides
local people with fuelwood, leaf litter, leaves for thatch, poles for construction,
honey, cattle fodder etc. all year round. Data for assessing the benefits of mangrove
afforestation has been collected from the residents of Dhulaswar village. Two FGDs
were conducted directly by the author. Because of local norms, females were reluctant
to speak publicly and hence the author interviewed a few of them individually. Apart
from this, several KIIs were conducted with local foresters, the Executive Engineer of
the Bangladesh Water Development Board (BWDB) and local community members.
During the focus group discussions and interviews it appeared that local people
are very aware regarding the benefits of local afforestation. All of them agreed that
Dhulaswar forest is an important resource for their village, and they concede with the
fact that this forest needs to be protected. Among the ecosystem services the forest
offers, local people mentioned that they are directly benefitted from the forest as they
collect fuelwood and leaves, use the forest as gazing ground and collect raw materials
to make mats. They also acknowledge that during adverse weather, the forest works
as a protective shield. In general, when the sea is rough, afforestation in the foreshore
reduces the impact of stormy winds in the hinterland. Local people could also recall
the service provided by the afforestation during cyclone Sidr. During FGDs, they
claimed that, during cyclone Sidr, the afforestation protected the embankment behind
it from surge waves. As a result, the embankment only breached where it is exposed
to the sea. They believe, if the afforestation did not exist, the damages would have
been much greater than it actually were.
4.7.1 Collected Goods and Uses
Among numerous extractable and non-extractable benefits provided by the small-

scale forests, in this study, dominant extractable benefits have been quantified. The
Forest Department prohibits the felling of whole trees but allows the collection of
non-main stem material such as branches, leaves and pneumatophores. Residents of
Dhulaswar village mainly collect firewood (mainly wood branches and dried leaves)
and Nipa palm (Locally known as Kewa pata). Nipa palm (Nypa fruticans) is easily
found in this forest and the adjacent area. During high tides and rainy season, the
forage area in the forest is considerably limited to local people. During low tides,
people can collect fuelwood from anywhere in the forest. As the forest is adjacent to
the embankment (used as village road), local people can easily access the forest on
foot. Moreover, as locals can extract regenerable mangrove species, they only need
permission from local foresters to enter the forest. According to the beneficiaries,
collected wood from this forest is not used for commercial purposes. All the families
use the collected fuelwood for cooking. Contrary to fuelwood, Nipa Palm is collected
to make mats, which are later sold at the village market.
4.7.2 Variation in Collection Patterns for Fuelwood
During data collection, local people were also requested to point out governing
parameters affecting their forage for goods throughout the year. Among several
issues, such as the distance of forest from home, safety, available work force, trans-
portation facilities and seasonal variation, it is found from the discussion that the
amount of goods collected from forest and the frequency, largely varies with the
distance of household from the forest, and also with the season. Detailed analyses of
these parameters are discussed below:
(i) Variation in collection pattern with distance.
Figure 8 represents the change in dependency on mangrove plantation for fuel-
wood as a function of distance. The figure is prepared from information collected
during FGDs with active participation of local people.
Fig. 8 Variation of dependency on forest for fuelwood with distance

In Fig. 8, the ellipsoids depict the level of dependency of local people on the forest
for combustible fuelwood. People living adjacent to the forest depend highly on the
forest and they forage almost the entire afforested area to collect wood (marked by the
red ellipsoid in the Fig. 8). There are about 60 families living in the red zone, and the
forest provides about 65% of total fuelwood consumed in a year. From the edge of the
forest, people living in about a strip of roughly 350 m–450 m are benefitted the most
from this afforestation of about 1.7 km2 (represented as red ellipsoid in the Fig. 8).
By the yellow ellipsoid, families of medium dependency are categorized along with
their forage distance. As estimated number of 80 families living in about 400 m
distance from the end of red zone fall in this category. Nearly 42% of annual needs
of residents in this zone are fulfilled with firewood from this mangrove forest. The
forage area of residents of this category is marked by yellow ellipsoid. Lastly, about
25 families living in a strip of 200 m–320 m (marked as green ellipsoid) is slightly
dependent on the forest for fuelwood. About 25% of required fuelwood is collected
from this forest. The remaining demand for firewood is fulfilled by non-mangrove
trees, which local businessmen sell at village markets. From the collected data, it can
be seen that as the distance between residence and forest increases, the dependency
level for fuelwood collection decreases significantly. During FGDs, local people
pointed out a number of reasons for this. First, there is no convenient local transport
to move collected fuelwood. People usually walk on foot towards their destination
after collecting wood, which hinders significant amount of fuelwood collection at
a time. As a result of this inconvenience, people living 1 to 1.5 km away from the
afforestation, prefer to buy their wood from the weekly village market. People living
outside the ellipsoids rarely collect goods from the forest, and hence they are excluded
from further data analysis.
(ii) Seasonal variation in the collection of fuelwood.
In addition to the distance, dependency of local people on the mangrove plantation
generally varies with the change in seasons. During FGDs, local people split an
entire year into two distinct categories considering the frequency of foraging and
the extracted quantity of firewood. These are summer and dry season. According
to them, summer season lasts from April to September, and dry season is from
October to March. During the FGDs, people estimated their average dependence
on the mangrove plantation for firewood by estimating the percentage of required
fuelwood collected in a month, frequency of foraging and the monthly extracted
quantity of firewood. The seasonal variation of three categories of users and related
data are represented in the Table 1.
As presented in Table 1, according to local people, people of high, medium and low
dependency extract more of their required fuelwood during dry season than summer.
During summer, dependence of people on the forest decreases significantly. For
people with medium and low dependency for fuelwood, an even sharper decrease
is observed. This is because during majority of the summer season, the weather is
very hot and humid, and sometimes, because of torrential rain, all the dried leaves
and dead branches are dampened, making them difficult to use. As a result, people
in general find it inconvenient to collect wet fuelwood. Especially, people living at
a significant distance from the forest, find it very difficult to forage. As mentioned
Table 1 Seasonal variation in the collection of fuelwood

Property Season Dependency Level
High Medium Low
Percentage of fuelwood collected Summer 40% 15% 5–7%
(monthly need) Dry 90% 70% 40%
Frequency of forage (Monthly) Summer 8–10 times 3–4 times 1–2 times
Dry At least 16 times 8–10 times 4–8 times
Monthly Extracted Quantity (in Summer 40–60 30–35 20
Kgs) Dry 90–120 60–85 Less than 40
* All the values presented in the table are approximated by local stakeholders
earlier, the road network of the village is also not good enough, so they are unable
to use any vehicles to move the fuelwood. So, people are not willing to go through
the trouble of collecting wet fuel, carrying them on foot and then drying them before
using in the summer season. So, they mostly prefer to buy fuel from the local market.
As a result, the frequency of forage drops largely during the summer. However, in the
winter or dry season, residents of the village, both men and women, are more willing
to collect fuelwood from the forest. As the weather is more convenient during the
winter, it is less tiresome for people to collect wood and carry it back to their home.
Since the weather is mostly sunny during the winter, people often collect slightly
wet wood and later dry it at home before using.
Figure 9 illustrates the seasonal benefit provided by the forest for three types of
users. As stated earlier, overall people get more benefits in the winter than in the
summer season. Though people with high dependency levels gets more benefit as
individual households, in this Figure, households under medium dependency show
a seasonal benefit of about Tk. 745.34 ($8.73) which is greater than for the high
dependency zone (Tk. 670.81/ $7.86). This is because the number of households in
this category is 20 greater than that of the high dependency ones. For summer season,
however, seasonal benefit remains inversely proportional to the distance, despite the
population in the medium dependency zone being considerably larger. Beneficiaries
with low dependency levels, show very little use of fuelwood (in monetary terms)
both in winter (Tk. 93.17/ $1.09) and summer (Tk. 46.58/$0.55).
4.7.3 Benefits Provided by Nipa Palm
Nipa Palm is mainly collected and used by village women to make mats which are
later sold in the weekly village market. It is mostly women from unprivileged families
who are engaged in making mats. From interviews with local women, it is found that
mats prepared from Nipa Palm make significant contributions to the monthly income
of about 20–25 families. On average, each family can make 10 mats a week. Later
these mats are sold at the village market by male members of the families. Presence
of afforestation has thereby created a source of income for underprivileged women.
800.00 745.34
670.81 ($8.74)
Seasonal Benefit Per hectare(In taka)
700.00 ($7.86)
600.00
500.00
335.40 winter
400.00 298.14
($3.93)
($3.50) Summer
300.00
200.00
93.17
($1.10) 46.58
100.00 ($0.55)
0.00
High Medium Low
Dependency Level
Fig. 9 Comparative change in seasonal benefit per hectare for fuelwood in three categories of user.
(1 USD = 85.34Taka, August 2021)
4.7.4 Net Benefit Per Hectare
Net benefit of Dhulaswar forest has been calculated using conceptual model (Eq. 4).
For this forest, locals claimed that there are no transportation costs (fixed or variable)
involved. Also, collectors never use fuelwood for commercial purposes. So, the
benefits provided by the forest for fuelwood is only the market value of the good,
which is Tk. 2.6 (US$0.03) per kilogram in this study. Each mat prepared from Nipa
palm is sold at about Tk. 120 (US$1.41) in the local market. The transportation cost
of carrying these mats to the market is slightly more than Tk. 100 (US$1.17) per
month. So, the benefit provided by Nipa Palm is calculated from subtracting the
transportation cost from the market price of the mats. Also, from FGDs with the
local forester, management cost of the forest (in this case provided by the Forest
Department) is found to be Tk. 9000 (US$105.46) per month. Lastly, the summation
of net benefits of each product is then divided by the area of afforestation which is
161 hectares. All these data are used in the conceptual model to get Table 2.
5 Conclusions
This study was conducted to examine the effectiveness of mangrove plantation in

reducing storm surge induced inundation, and in the provision of services to local
people at polder 47/4 in Kalapara Upazila of Patuakhali district in Bangladesh. A
Table 2 Calculation of net benefit per hectare

Products Benefit Per year Variable cost per Management cost Net
(in taka) year in taka per year (in taka) benefit/hectare/year
(in Taka)
Fuel wood 3,52,500 0 1,08,000 5990.68 ($70.20)
($4130.53) ($1265.52)
Nipa Palm 69,600 ($815.56) 2,400 ($28.12)
validated coastal model of Bangladesh was used to simulate and analyze how the pres-
ence of small-scale plantation can have a positive impact in reducing flood inundation
during cyclone Sidr. For Dhulaswar reserved forest in polder 47/4, existence of this
forest has been found to have saved an additional 3 km2 from coastal flooding during
cyclone Sidr. Comparing conventional coast protection approaches with hybrid inter-
vention, it is found that providing plantation to offer predetermined resistance can be
integrated as a feasible alternative, with existing polder height in lieu of increasing
polder height in places where land is available or because of local hydrodynamics
during cyclone, a certain portion of polder poses more threat to breach.
Plantation also offers economic benefits to local people. In case of polder 47/4,
services provided by the mangrove ecosystem vary with the distance of the forest
from households, and there is also a seasonal variation in terms of foraging. These
can be important factors in determining the composition of vegetation species for
the coastal green belt, especially if the plantation project is designed as participatory
forestry. In this study, direct extractable benefits from Dhulaswar forest are quantified
to be Tk. 5990.68 (US$70.20) per hectare per year.
Appendix
Model Domain
Validation of Model for Cyclone Sidr
From storm surge validation, trial runs are executed in Delft-Dashboard. The simu-
lations also include tidal constituents from open boundary for both events. For the
Sidr 2007 cyclone, Indian Meteorological Department (IMD) estimated maximum
wind speeds of about 60 m/s during landfall and the lowest central pressure of the
cyclone to be 944 hPa with a pressure drop of 66 hPa during landfall. The couple
model run was performed for the entire cyclone period of 5 days from its generation
at 0000 UTC, 11 November 2007, to its dissipation on 0000 UTC, 16 November
2007.
The model results were observed to perform well while calculating the inundation
extent for S 2007 (Fig. 11). High water-level values were also extracted for the same
Fig. 10 Model domain along with upstream and downstream boundary of the model. (Al Azad
et al. 2018)
locations (Table 3) from previous studies conducted by JSCE, 2008b, Lewis et al.
2013 and Deb and Ferreira, 2017 (Table 3). The computed storm surge was found to
vary between 2.0 and 6.0 m along the coastal districts and estuaries where Sidr 2007
made landfall, and previous studies also demonstrated similar results on the same
locations (Table 3).
Fig. 11 Inundation map of cyclone Sidr. (Haque et al., 2018)
Table 3 Comparison of model outcomes with high water levels (HWLs) extracted from existing
studies on storm surge modelling in Bangladesh coast for cyclone Sidr
Location Hiron Point Khepupara Chittagong Cox’s Bazar
Coordinates 21.81 21.88 22.19 21.46
89.49 90.10 91.92 91.92
Model data 2.31 5.86 3.12 2.1
JSCE,2008 2.2 5.9 3.0 2.0
% of error (between Model data and −5% 0.68% −4% −5%
JSCE 2008)
Lewis et al. (2013) 2.5 5.5 3.2 2.0
% of error (between model data and 7.6% −6.55% 2.5% −5%
Lewis et al., 2013)
Deb and Ferreira (2017) 2.5 6.0 3.0 1.7
% of error (between model data and Deb 7.6% 2.33% −4% −23.53%
and Fererra 2016)
Sample Area for Manning’s Roughness Calculation
Fig. 12 Manning’s roughness calculation sample area (a) Section-1: less dense area (represented
by orange box) and another one is (b) Section-2: comparatively denser area (represented by the blue
box)
Polder Height Measurement
Table 4 Measured dimensions of polder at selected locations

Location GPS coordinate Height (m) Width(m) Landward slope Seaward slope
(in decimal (degree) (degree)
degree))
Point 1 21.8479, 90.2453 4.8 4.8 55 42
point 2 21.8413, 90.2385 5.43 4.01 60 44
Point 3 21.8443, 90.2192 3 4.42 69 50
Fig. 13 Photos of locals

collecting and using
firewood and nipa plam from
the forest (Source Author)
Fig. 13 (continued)
Fig. 13 (continued)
Fig. 13 (continued)
References
Abdullah, Abu Nasar, Natasha Stacey Mohammad, Stephen T. Garnett, and Bronwyn Myers.
2016. Economic Dependence on Mangrove Forest Resources for Livelihoods in the Sundarbans,
Bangladesh. Forest Policy and Economics 64: 15–24. https://doi.org/10.1016/j.forpol.2015.12.00.
Al Azad, A. S. M. Alauddin, Kazi Samsunnahar Mita, Md Zaman, Marin Akter, Tansir Zaman
Asik, Anisul Haque, Mohammad Asad Hussain, and Md Rahman. 2018. Impact of tidal phase
on inundation and thrust force due to Storm Surge. Journal of Marine Science and Engineering
6 (4): 110. https://doi.org/10.3390/jmse6040110.
Arcement Jr, G. J., & Verne R. Schneider. 1989. Guide for Selecting Manning’s Roughness Coef-
ficients for Natural Channels and Flood Plains. United States Geological Survey Water-supply
Paper 2339. https://doi.org/10.3133/wsp2339.
Arkema, Katie K., Greg Guannel, Gregory Verutes, Spencer A. Wood, Anne Guerry, Mary Ruck-
elshaus, Peter Kareiva, Martin Lacayo, and Jessica M. Silver. 2013. Coastal Habitats Shield
People and Property from Sea-Level Rise and Storms. Nature Climate Change 3 (10): 913–918.
https://doi.org/10.1038/nclimate1944.
Beck, Michael W., and C. Shepard. 2012. Coastal Habitats and Risk Reduction. World Risk Report.
Berlin: Bundnis Entwicklung Hilft.
Bridges, Todd, R. Henn, S. Komlos, D. Scerno, T. Wamsley, and K. White. 2013. Coastal
Risk Reduction and Resilience: Using the Full Array of Measures. Washington, DC. http://sag
ecoast.org/docs/sci_eng/USACE_Coastal_Risk_Reduction_final_CWTS_2013-3.pdf. Accessed
June 27, 2021.
CEGIS (Center for Environmental and Geographic Information Services). 2016. Final Report of
Technical Study for Mapping of Potential Greenbelt Zone in the Coastal Regions of Bangladesh.
Dhaka: CEGIS.
Cheong, So-Min, Brian Silliman, Poh Poh Wong, Bregje Van Wesenbeeck, Choong-Ki Kim, and
Greg Guannel. 2013. Coastal Adaptation with Ecological Engineering. Nature Climate Change
3 (9): 787–791. https://doi.org/10.1038/nclimate1854.
Chow, Jeffrey. 2015. Spatially Explicit Evaluation of Local Extractive Benefits from Mangrove
Plantations in Bangladesh. Journal of Sustainable Forestry 34 (6–7): 651–681. https://doi.org/
10.1080/10549811.2015.1036454.
Chow, Ven T. 1959. Open-Channel Hydraulics. McGraw-Hill Civil Engineering Series.
Chowdhury, A., R. Mushtaque, Abbas U. Bhuyia, A. Yusuf Choudhury, and Rita Sen. 1993. The
Bangladesh Cyclone of 1991: Why So Many People Died. Disasters 17 (4): 291–304. https://doi.
org/10.1111/j.1467-7717.1993.tb00503.x.
Das, Saudamini, and Jeffrey R. Vincent. 2009. Mangroves Protected Villages and Reduced Death
Toll During Indian Super Cyclone. Proceedings of the National Academy of Sciences 106 (18):
7357–7360. https://doi.org/10.1073/pnas.0810440106.
Dasgupta, Susmita; Huq, Mainul; Khan, Zahirul Huq; Ahmed, Manjur Murshed Zahid; Mukherjee,
Nandan; Khan, Malik Fida; Pandey, Kiran. 2010. Vulnerability of Bangladesh to Cyclones in a
Changing Climate: Potential Damages and Adaptation Cost. Policy Research Working Paper no.
WPS 5280. World Bank. Last accessed on June 26, 2021. Available online at https://openknowl
edge.worldbank.org/handle/10986/3767.
De Vriend, Huib J., Mark van Koningsveld, Stefan GJ Aarninkhof, Mindert B. de Vries, and Martin
J. Baptist. 2015. Sustainable Hydraulic Engineering Through Building with Nature. Journal of
Hydro-environment research 9 (2): 159–171. https://doi.org/10.1016/j.jher.2014.06.004.
Deb, M., and C.M. Ferreira. 2017. Potential Impacts of the Sunderban Mangrove Degradation on
Future Coastal Flooding in Bangladesh. Journal of Hydro-Environment Research 17: 30–46.
Ferrario, Filippo, Michael W. Beck, Curt D. Storlazzi, Fiorenza Micheli, Christine C. Shepard, and
Laura Airoldi. 2014. The Effectiveness of Coral Reefs for Coastal Hazard Risk Reduction and
Adaptation. Nature Communications 5 (1): 1–9. https://doi.org/10.1038/ncomms4794.
GoB 2008. Cyclone Sidr in Bangladesh Damage, Loss, and Needs Assessment for Disaster Recovery
and Reconstruction. Dhaka: Government of Bangladesh assisted by International Development
Community with Financial Support from the European Commission.
GoB. 2013. Disaster Report. Department of Disaster Management, Ministry of Disaster Manage-
ment and Relief: Government of the People’s Republic of Bangladesh.
Hydraulics, WL Delft. 2006. Delft3D-FLOW: Simulation of Multidimensional Hydrodynamic Flows
and Transport Phenomena, Including Sediments, User Manual. Delft, Holland: Deltares.
India Meteorological Department (IMD). 2021. Ministry of Earth Sciences, Government of India.
Available online: http://imd.gov.in. Accessed June 27, 2021.
Islam, Md Nazrul, Nabila Hasan Dana, Khandkar-Siddikur Rahman, Md Tanvir Hossain, Moin
Uddin Ahmed, and Abdulla Sadig. 2019. Nypa Fruticans Wurmb Leaf Collection as a Livelihoods
Strategy: A Case Study in the Sundarbans Impact Zone of Bangladesh. Environment, Development
and Sustainability, 1–18. https://doi.org/10.1007/s10668-019-00438-w.
Jisan, Mansur Ali, Shaowu Bao, and Leonard J. Pietrafesa. 2018. Ensemble Projection of the Sea
Level Rise Impact on Storm Surge and Inundation at the Coast of Bangladesh. Natural Hazards
and Earth System Sciences 18 (1): 351–364. https://doi.org/10.5194/nhess-18-351-201.
Jones, Holly P., David G. Hole, and Erika S. Zavaleta. 2012. Harnessing Nature to Help People
Adapt to Climate Change. Nature Climate Change 2 (7): 504–509. https://doi.org/10.1038/ncl
imate1463.
JSCE. 2008a. Investigation Report on the Storm Surge Disaster by Cyclone Sidr 2007. Investigation
Team of Japan Society of Civil Engineering, Bangladesh. Available at: http://www.jsce.or.jp/rep
ort/46/files/Bangladesh_Investigation.pdf. Accessed 27 June, 2021.
JSCE. 2008b. Investigation Report on the Storm Surge Disaster by Cyclone Sidr 2007. Investigation
Team of Japan Society of Civil Engineering, Bangladesh. Available at: http://www.jsce.or.jp/rep
ort/46/files/Bangladesh_Investigation.pdf. Retrieved at November 2018.
Lewis, M., P. Bates, K.J. Horsburgh, J. Neal, and G. Schumann. 2013. A Storm Surge Inundation
Model of the Northern Bay of Bengal Using Publicly Available Data. Quarterly Journal of the
Royal Meteorological Society 139: 358–369.
Mita, Kazi Samsunnahar. 2019. Role of Mangrove Plantation in Improving Embankment Safety
and Supporting Local Livelihoods. M.Sc. Thesis. Bangladesh University of Engineering and
Technology.
Needham, Hal, and Barry D. Keim. 2011. Storm Surge: Physical Processes and an Impact Scale.
Recent Hurricane Research-Climate, Dynamics, and Societal Impacts. Anthony Lupo, Inte-
chOpen. Available online: https://www.intechopen.com/download/pdf/15338. Accessed August
28, 2021. https://doi.org/10.5772/15925.
NYS 2100 Commission. 2013. Recommendations to improve the strength and resilience of the
Empire State’s Infrastructure. New York State Official Report, Andrew Cuomo, Governor. last
accessed on June 26, 2021. http://www.governor.ny.gov/assets/documents/NYS2100.pdf.
Quadir, Dewan Abdul. 2008. Tropical Cyclones: Impact on Coastal Livelihoods: Investigation of
the Coastal Inhabitants of Bangladesh. Dhaka: IUCN Bangladesh Country Office.
Reddy, Sheila MW., Gregory Guannel, Robert Griffin, Joe Faries, Timothy Boucher, Michael
Thompson, Jorge Brenner, et al. 2016. Evaluating the Role of Coastal Habitats and Sea-Level Rise
in Hurricane Risk Mitigation: An Ecological Economic Assessment Method and Application to
a Business Decision. Integrated Environmental Assessment and Management 12 (2): 328–344.
https://doi.org/10.1002/ieam.1678.
Santiago Fink, Helen. 2016. Human-Nature for Climate Action: Nature-Based Solutions for Urban
Sustainability. Sustainability 8 (3): 254. https://doi.org/10.3390/su8030254.
Simm, Jonathan, Amy Guise, David Robbins, and Jason Engle. 2016. US North Atlantic Coast
Comprehensive Study: Resilient Adaptation to Increasing Risk. In Coastal Management:
Changing Coast, Changing Climate, Changing Minds, 639–649. ICE Publishing. https://doi.
org/10.1680/cm.61149.639.
slam, M.N., Dana, N.H., Rahman, KS. et al. 2020. Nypa Fruticans Wurmb Leaf Collection as a
Livelihoods Strategy: A Case Study in the Sundarbans Impact Zone of Bangladesh. Environ Dev
Sustain 22, 5553–5570. https://doi.org/10.1007/s10668-019-00438-w.
Small, Christopher, and Robert J. Nicholls. 2003. A Global Analysis of Human Settlement in Coastal
Zones. Journal of Coastal Research, 584–599. Accessed on June 27, 2021. http://www.jstor.org/
stable/4299200.
Sutton-Grier, Ariana E., Kateryna Wowk, and Holly Bamford. 2015. Future of Our Coasts: The
Potential for Natural and Hybrid Infrastructure to Enhance the Resilience of Our Coastal Commu-
nities, Economies and Ecosystems. Environmental Science & Policy 51: 137–148. https://doi.org/
10.1016/j.envsci.2015.04.006.
UNDP. 2015. Integrating Community-Based Adaptation into Afforestation and Reforestation
Programmes in Bangladesh. Government of Bangladesh, Bangladesh Forest Department, Dhaka.
Heerden, Van, and Ivor Ll. 2007. The Failure of the New Orleans Levee System Following Hurricane
Katrina and the Pathway Forward. Public Administration Review 67: 24–35. https://doi.org/10.
1111/j.1540-6210.2007.00810.x.
Van Slobbe, E., J. de Huib, S. Vriend, K. Lulofs. Aarninkhof, Marjan de Vries, and P. Dircke. 2013.
Building with Nature: In Search of Resilient Storm Surge Protection Strategies. Natural Hazards
66 (3): 1461–1480. https://doi.org/10.1007/s11069-012-0342-y.
Zhang, Keqi, Huiqing Liu, Yuepeng Li, Hongzhou Xu, Jian Shen, Jamie Rhome, and Thomas J.
Smith III. 2012. The Role of Mangroves in Attenuating Storm Surges. Estuarine, Coastal and
Shelf Science 102: 11–23. https://doi.org/10.1016/j.ecss.2012.02.021.
Assessing the Consequences
of Large-Scale Stabilization of the Padma
River on Its Flow Hydraulics Using
a Combined 1D-2D Hydrodynamic
Model
Subir Biswas and M. Shahjahan Mondal
Abstract Bangladesh is in a unique geographic location dominated by floodplains

and consists of three large and extremely dynamic rivers—the Ganges, the Brahma-
putra, and the Meghna. The Government of Bangladesh has policies and plans
regarding water resource management, disaster risk reduction, and land reclamation
from the riverbed. Such policies and plans include narrowing down, channelizing
and stabilizing the Padma River, and constructing structures on its floodplains. The
rehabilitation and construction of embankments along the river have been suggested
under the stabilization programs. However, such stabilization may result in a change
of hydraulic characteristics of the Padma River. Very little studies have been done to
assess the effects of such large-scale stabilization on the hydraulic behavior of the
river. This study has set up a coupled 1D-2D hydrodynamic model using HEC-RAS
for the Padma River system and simulated the potential impacts of such stabilization.
The Padma River, its distributaries, and adjoining areas are included in the model
setup. The model is calibrated for the 1998 flood and validated for the 2004 flood
using the observed water level. Further calibration and validation at the floodplain
have been done using the primary highest flood level data, and through a qualitative
comparison of flood inundation with satellite images, respectively. It has been found
that the model can simulate the observed variation of these floods quite satisfactorily.
The simulation results reveal that the peak discharge of the 1998 flood can increase
by 20% in the Padma River under a stabilized condition. The peak flood level in
the river can increase by 51 cm, and the maximum average velocity by 25% due to
the stabilization. The outcomes of this study will be useful in training the river and
designing other interventions for river management.
Keywords River stabilization · Padma River · Embankment · River hydraulics ·

1D-2D-coupled HEC-RAS model
S. Biswas (B) · M. S. Mondal

Technology (BUET), Dhaka, Bangladesh
280 S. Biswas and M. S. Mondal
1 Introduction
Bangladesh is located at the most downstream of the world’s largest delta. It

consists of floodplains of the Ganges, Brahmaputra, and Meghna rivers, and their
numerous tributary and distributary rivers. About 80% area of the country consists
of the combined delta of these three international rivers (Islam 2011). This unique
geographic location makes Bangladesh highly vulnerable to flooding. In general, four
types of floods occur in this country: flash floods, riverine floods, tidal floods, and
floods due to storm surge. Riverine floods are the most common and such floods occur
in many parts of the country almost every year. Heavy monsoon rainfall and synchro-
nization of flood-peaks of the major rivers are generally considered to be the main
causes of this flood. Normally, during the monsoon season, 25–30% of the area is
inundated, but during extreme flood events, 50–70% area of the country can be subject
to inundation (Hossain 2003). The floods of the years 1988 and 1998 inundated about
61% and 68% area of the country, respectively (FFWC 2018). Floods cause signif-
icant damage to crops, lives, and property every year. The 1998 flood damaged an
estimated 15,000 km of roads, 14,000 schools, over 500,000 homes, thousands of
bridges and culverts, and 2.04 million metric tons of rice crop production (Ninno
et al. 2001).
The Government of Bangladesh (GoB) has undertaken many policies, plans, and
projects for better water resource management and disaster risk reduction. Such poli-
cies, plans, and projects (e.g., Flood Control, Drainage and Irrigation Projects, Flood
and Riverbank Erosion Risk Management Investment Programs) include channel-
ization and stabilization of the major rivers and reclamation of lands from the river
beds to increase economic activities on the floodplains (NHC 2013, BWDB 2014,
GED 2017). It has become a common practice to construct earthen embankments
(also known as levees or dykes in some countries) along the rivers to confine the river
flows within the river cross-sectional areas, and to thereby prevent flooding of the
floodplains. During the 1960s, many earthen embankments were constructed along
the major rivers for flood protection. The GoB has constructed a total of 16,261 km
of embankments including 5757 km of coastal embankments (BWDB 2019). Simi-
larly, the ongoing and future initiatives of GoB for water resource management and
disaster risk reduction indicate that the Padma River will be narrowed down to a single
channel and stabilized through engineering works. The average width of the Padma
River will be reduced to 5.3 km from 10.6 km by the year 2030 through channelization
and riverbank stabilization (NHC 2013). There will be embankments on both sides
of the river to protect life and property from flooding. However, such flood-control
embankments and engineering works may result in a change of hydraulic character-
istics of the river, which may have both positive and negative impacts (Baird et al.
2015). Also, to design stabilization structures like flood-control embankments along
the river, necessary analysis must be conducted to understand the future behavior of
the river (e.g., changed discharge, water level, and velocity) under the proposed stabi-
lized condition. This stabilization impact can be determined through hydrodynamic
model simulations. However, very little studies have been done to assess the effects
Assessing the Consequences of Large-Scale Stabilization … 281
of large-scale stabilization on the hydraulic behavior of the rivers. In this study, the
impact of large-scale stabilization of the Padma River on its discharge, water level,
and velocity magnitude is assessed through a combined one- and two-dimensional
(1D-2D) HEC-RAS model simulation for the Padma River system. Among the avail-
able models for hydraulic simulation, the HEC-RAS, developed by the U.S. Army
Corps of Engineers, provides automated floodplain mapping and analysis with effec-
tive, proficient, and standard results. The combined 1D-2D HEC-RAS model is more
suitable for a large river network having a floodplain, where a more detailed flow
field is necessary (Brunner 2014).
2 Study Area and Methodology
Considering the strategies of the Bangladesh Delta Plan 2100 about river stabilization,
this study investigates similar stabilization impacts on the Padma River, which is one
of the three major rivers in Bangladesh. The River Ganges confluences with the
Brahmaputra-Jamuna River at Goalundo and holds the name as the Padma River.
This part of the river mainly has meandering characteristics, but in some reaches, it
shows similarity to the braided and straight patterns both spatially and temporally
(McLean et al. 2012). At Chandpur, the Padma River joins the Upper Meghna River
and holds the name as the Lower Meghna River and flows to the Bay of Bengal
(BWDB 2011). Due to its connectivity with the sea via the Lower Meghna, the
Padma River experiences tidal influence. The Padma River and its floodplains have
been considered as the study area in this study (Fig. 1).
The Padma River has a total length of about 120 km. The length of the river consid-
ered in this study is about 94 km, and the total area of the floodplain is about 2843 km2 .
Required data for this study have been collected from primary and secondary sources.
Secondary data include water level, discharge, and river bathymetry, digital elevation
model (DEM), satellite images, and others. Some of these data were already available
with the Institute of Water and Flood Management (IWFM), and others were collected
from the Water Resources Planning Organization (WARPO) and Bangladesh Water
Development Board (BWDB). Primary data were collected during the field visits.
A 1D-2D-coupled HEC-RAS model for the Padma River System was set up. The
model was calibrated with the measured data for the 1998 flood and validated inde-
pendently with the measured data of the 2004 flood. After calibration and validation
of the model, embankments on both banks of the Padma River were added using
lateral structure along the river. Afterward, the model was rerun, and the results were
analyzed through a comparison with the measured data. For the impact analysis,
the 1998 flood was chosen as it is historically the highest flood in the Padma River
floodplain.
Fig. 1 Study Area shown on the Google Earth map
2.1 Development of 1D-2D-Coupled HEC-RAS Model
As the study area consists of a deltaic floodplain, its topography is moderately flat.
The Padma River also has tidal influences. The water from the river can easily enter
the floodplain during high tide and depart the floodplain during low tide, especially in
the monsoon season (Mondal et al. 2018). These floodplains act as storage reservoirs
and play an important role in river hydraulics. So, a combined 1D-2D model was
developed to simulate the flooding pattern where the rivers are modeled as the 1D
elements and the floodplains are modeled as the 2D elements.
First, the 1D river network of the study area has been developed in the Geometric
Data of the HEC-RAS model. The Padma River and its distributaries have been
delineated using the River Reach tool following the centerline and deepest channel
as a guide. After delineation of the river network, the bathymetries of the rivers were
entered into the HEC-RAS Geometry. In the network, the Padma River, as well as
its distributaries (the Arial Khan, the Kirtinasa, and the Joynagar-Ganganagar), is
included. The current model setup covers a total length of about 262 km of river
network including the Padma River and its distributaries as 1D element. Of these,
the Padma River itself covers a length of about 94 km. For an accurate determination
of the energy gradient, the measured cross-sections have been interpolated using
Fig. 2 A schematic showing

the river network used in the
coupled 1D-2D HEC-RAS
model
additional master cords. A schematic showing the river network used in the model
is given in Fig. 2.
After developing the 1D river network, the 2D flow area is added by drawing a 2D
flow area polygon and developing the 2D computational mesh. For the development
of the 2D floodplain, a terrain model is created in the HEC-RAS and is then added
in the RAS Mapper. The resolution of the terrain model is a limiting factor for
determining the quality of the hydraulic model. A high-resolution terrain model is
essential to create a detailed and accurate hydraulic model. A DEM with a 500
m resolution from WARPO has been used to create the terrain model. The DEM
resolution was further improved by using the topographical survey data collected
in some selected areas by IWFM (2018). The raster DEM was saved in a float file
format as required in the HEC-RAS. The raster file of the study area has been added
to the RAS Mapper by using the new terrain layer tool. The raster file has been
converted into the GeoTIFF (*.tif) file format by RAS Mapper (Brunner 2016). The
terrain model has been further improved by using the cross-section data in the RAS
Mapper. The terrain model of the study area thus created is shown in Fig. 3.
The terrain file is then opened in the Geometric Data Window of the HEC-RAS for
creating a 2D flow area. A 2D flow area is added by drawing a 2D flow area polygon
boundary in the Geometric Data Window. Due to the flat topography, the river water
can easily exchange between the river and the floodplain with changes in the stages.
Nowadays, the floodplains in many places are obstructed by highways, roads, and
flood-control embankments. So, only the unobstructed parts of the floodplain act as
storage areas. Such barriers and terrain contour lines were used as a guide to select
the 2D area extent, as suggested in the HEC-RAS User Manual (Brunner 2016). It
is found through several unsteady flow simulations, that a little change in the extent
of 2D floodplain areas has no impact on the model outputs such as maximum water
Fig. 3 Terrain model

developed for the study area
level and velocity (Biswas et al. 2020). The number of 2D areas in the model is
six, and the total area is about 2843 km2 . After drawing the 2D flow area polygon
boundary, the 2D computational mesh and hydraulic property tables for the 2D cells
and cell faces have been created. A 100 m × 100 m grid resolution has been used
for the computational mesh.
After developing the 2D computational mesh, it has been connected with the 1D
river using lateral structures from the Geometry Data Window. Here, a series of weir
structures on each riverbank has been created to establish the connectivity between
the 1D river and the 2D flow area.
2.2 Boundary Conditions
The model setup used in this study is for an unsteady flow simulation and thus requires
the boundary and initial conditions. There is one upstream discharge boundary, two
downstream water level boundaries, and five 2D area peripheral boundaries which
act as the outflow boundary conditions. The locations of the boundaries are shown in
Fig. 4. Here, the green arrow shows the upstream boundary, the two yellow arrows
show the two downstream water level boundaries, and the red lines show the five
peripheral boundaries.
The upstream boundary of this model is at Baruria Transit (SW91.9L) on the
Padma River where the mean daily discharge (MDD) has been used. The two down-
stream boundaries are on the Padma River and on the Arial Khan River. The water
levels at these boundaries for the year 1998 are obtained from a 1D hydrodynamic
Fig. 4 Locations of the

boundaries in the HEC-RAS
model
HEC-RAS model developed by IWFM (2018) and Mondal et al. (2018) which was
well-calibrated for the 2014 flood and validated for the 1998 and 1988 floods. Since
the two downstream boundaries experience tidal influence, three hourly water level
has been used as the boundary conditions at the downstream. For the year 2004, for
both the downstream boundary conditions, such as Madaripur (SW5) on the Arial
Khan and Sureswar (SW95) on the Padma River, the water levels were used after
slope analysis as they are located further downstream. Observed discharge at Baruria
Transit and water level on the Padma and Arial Khan were used as an initial condi-
tion. The peripheral outflow boundaries are modeled with normal depth. The normal
depth boundary conditions can be used at locations where the flow leaves a 2D flow
area or at an open-ended reach. Here, the peripheral boundaries are located on the
floodplain where measured data (e.g., water level) are not available. So, the normal
depth is used as recommended by Brunner (2016). Roy (2019) also used a similar
approach. It requires entering a friction slope of that area to use the normal depth
as a boundary condition. The friction slope should be based on the land slope at the
vicinity of the 2D flow area boundary condition line. This friction slope is used in
Manning’s equation to compute a normal depth for each given flow, based on the
cross-section underneath the 2D boundary condition line. In this model, there are
five outflow boundaries where the average bed slope at the vicinity of the boundary
has been used as an estimate of the friction slope. Two of them are located near the
Padma River where a friction slope of 0.0002 has been used, and for others located
near the Arial Khan River and open floodplain, a friction slope of 0.0001 has been
used.
2.3 Model Simulations
After completing the model setup, the unsteady flow simulations were performed
from 4 June to 22 September for the years of 1998 and 2004, to estimate the flood-
peaks with a computational time interval of one minute. HEC-RAS solves either
the full 2D Saint Venant equations or the 2D Diffusion Wave equations during the
unsteady flow simulation. The governing equations for surface flow can be written
as below:
∂ζ ∂ p ∂q
+ + =0 (1)
∂t ∂x ∂y

∂p ∂ p2 ∂ pq n 2 pg p 2 + q 2 ∂ζ
+ + =− − gh + pf
∂t ∂x h ∂y h h2 ∂x
∂ ∂
+ (hτx x ) + hτx y (2)
ρ∂ x ρ∂ y

∂q ∂ q2 ∂ pq n 2 qg p 2 + q 2 ∂ζ
+ + =− − gh +qf
∂t ∂y h ∂x h h2 ∂y
∂ ∂
+ hτ yy + hτx y (3)
ρ∂ y ρ∂ x
where, ζ = surface elevation, p = specific flow in the x direction, q = specific flow

in the y direction, h = water depth, g = gravitational acceleration, n = Manning’s
resistance coefficient, f = Coriolis, ρ = water density, τ = components of the
effective shear stress.
Equation 1 is the continuity equation and Eqs. 2 and 3 are the momentum equa-
tions. The acceleration terms in the Saint Venant equations can be neglected in most
practical applications and then these become simple parabolic equations, known as
the diffusive wave equations (Ponce 1990).
The full 2D Saint Venant equations are sometimes referred to as the full 2D
shallow water equations. The user can select either of the equations. In general,
the 2D Diffusion Wave equations allow the model to run faster and have greater
stability properties. Quiroga et al. (2016) considered both options, the full Saint
Venant equations and the 2D diffusion wave equations. They found that both methods
provide the same results, but simulation solving the 2D diffusion wave equations was
about 20 times faster. In this study, Diffusion Wave equations have been used as this
model simulates only the monsoon flood.
As most of the DEMs used in the HEC-RAS model are from the year 2001, the year
1998 was chosen for the model calibration, and the year 2004 for the model validation.
Also, the highest flood in the study area was observed in 1998, and there was a
moderate flood in 2004. Using the boundary and initial conditions for each year, the
model was run and the computed stage hydrographs were compared with the observed
stage hydrographs to judge the performance of the model. As the model calculates
water level at all points within the model domain, the simulated water level has been
compared with the Bhagyakul gage (SW 93.4L) station’s observed water level to
calibrate and validate the model. Bhagyakul is an intermediate gage station on the
Padma River and is located near the center of the model domain. Manning’s roughness
coefficients for the main channel and floodplain were the principal calibration and
validation parameters. The final calibrated and validated roughness parameters are
found to be higher for the overbank areas (n = 0.030) and lower for the main channels
(n = 0.025) and the highest in the floodplain (n = 0.05–0.06). Calibration at the 2D
floodplain for 1998 has been done by comparing the survey data on the highest flood
level and model-simulated highest water level for the same location. Validation at
the 2D floodplain for 2004 has been done by a qualitative comparison between the
MODIS image and the model-simulated flood inundation map. A similar approach
for comparison of the 2D model flood inundation with the available satellite images
has been used by many apprentices (Nishat 2017; Roy 2019; Tazin 2018).
There are many ways to evaluate a model’s performance. Moriasi et al. (2007)
have proposed guidelines for evaluation of model performance and recommended
model evaluation techniques. They recommended quantitative statistics with graph-
ical techniques. They divided quantitative statistical techniques for model perfor-
mance evaluation into three major categories: standard regression, dimensionless,
and error-index. In this study, three extensively used quantitative statistical perfor-
mance indicators, namely Coefficient of Determination (R2 ) as standard regression,
Coefficient of Nash–Sutcliffe Efficiency (NSE) as dimensionless, and Root Mean
Square Error (RMSE) as error-index, have been used for comparison of the simulated
daily average water level with the observed daily average water level.
The embankments have been modeled using lateral structures along the river.
There is an embankment on the left bank of the Padma River (Dhaka Southwest
Project Embankment). The embankments were initially built with set-back distances
from the active channel of around 1.5 km (NHC 2013). This distance has been
followed as a setback to locate the embankment (lateral structure) in this model,
with a few changes in some critical areas. The height of the embankment has been
set in such a way that water cannot enter the floodplain. So, all the water remains
within the river cross-sectional area for simulation under such a stabilized condition.
Afterward, the results have been analyzed.
3 Results
In this 1D-2D-coupled model simulation, a spatially variable Manning’s roughness

coefficients were the principal calibration parameters. The calibration and validation
results for the 1D river and 2D floodplains are discussed below.
3.1 Calibration and Validation Results
The model-simulated water level for Bhagyakul (SW 93.4L) on the Padma River is
shown in Fig. 5 along with the observed water level at the same station. The figure
shows that the water level has been simulated very well, except at the beginning.
The inconsistency at the beginning could be due to the warm-up time required for an
unsteady flow model to become stable. Elsewhere, the model captures the variation
in the stage hydrograph at Bhagyakul reasonably well.
The highest measured water level at Bhagyakul was found to be 7.50 mPWD on 10
September and the highest simulated water level was 7.57 mPWD on the same day.
Thus, the model-simulated water level was only about 7 cm higher than the measured
water level at the peak. Also, it is seen from the figure that, the model could capture
the variation in major part of the monsoon hydrograph very well. Since this study is
mainly concerned with the peak flood period, the simulation results can be considered
quite satisfactory. Since we are simulating the water level of 1998 using recent river
bathymetries, and the river network used is very large, this small discrepancy could be
expected and the model result can be considered to be acceptable. During calibration,
the values of R2 and NSE were found to be 0.98 and 0.95, respectively. These values
indicate that the simulated water level is very close to the observed water level and
also within the satisfactory range. The RMSE was found to be 0.20 m.
The validation of the model has been done for the year 2004. The comparison
of the observed and simulated discharge hydrographs at Bhagyakul (SW 93.4L) is
shown in Fig. 6.
It is seen from the figure that, the model captures the variation in major part
of the monsoon hydrograph very well, with some discrepancy at the beginning of
the hydrograph like the calibration year 1998. However, the peak does not match
precisely in this case. This may be due to the unexpected drops in the observed
discharges before the peak (on 15 and 22 July 2004) at Baruria Transit which was
used as the upstream boundary. Otherwise, the simulation result can be considered
quite satisfactory. In the validation, the values of R2 and NSE have been found to
be 0.98 and 0.90, respectively, which indicate that the simulated water level is very
Fig. 5 Comparison of 8.0

simulated and observed
water levels at Bhagyakul on 7.0
Water Level (mPWD)
the Padma River for 1998

6.0
5.0
Observed
Model
4.0
3.0
4/Jun/98 24/Jun/98 14/Jul/98 3/Aug/98 23/Aug/98 12/Sep/98

simulated and observed
water levels at Bhagyakul on 7.0
Water Level (mPWD)

the Padma River for 2004
6.0
5.0
Observed
Model
4.0
3.0
close to the observed water level and also within the satisfactory range. The RMSE
was found to be 0.27 m.
3.2 Comparison of Flooding at 2D Floodplain
A comparison of flooding at 2D floodplain has been done through the comparison

of the highest flood levels from the field survey, and the model simulation for the
1998 flood. The survey was conducted by IWFM during April–June of 2017, on the
maximum flood level of the 1998 flood in Shariatpur and Madaripur districts covering
an area of about 80 km2 . Those surveyed points are within this model domain. A
relation between the surveyed maximum flood level and the modeled maximum flood
level is shown in Fig. 7.
It is found that the model-simulated maximum flood level is 19 cm higher on
average than the survey data, and the highest difference is 55 cm. The values of
R2 and NSE have been found to be 0.85 and 0.61, respectively, which fall within a
satisfactory range. The RMSE was found to be 0.25 m.
A comparison of flooding at 2D floodplain has been done by comparison with
the available satellite images for the 2004 flood. The qualitative comparison has
been done for two situations—before flood conditions and during flood conditions.
Figures 8 and 9 show the comparisons between flood inundation maps from MODIS,
and that from the model for both before the flood and during the flood conditions.

Maximum Flood Level (mPWD)
maximum flood levels 6.8 R² = 0.8476

between field survey and 6.6
model simulation for 1998
from Model
6.4
flood 6.2
6.0
5.8
5.6
5.4
5.4 5.6 5.8 6.0 6.2 6.4 6.6 6.8 7.0
Maximum Flood Level (mPWD) from Field Survey
Fig. 8 Before-flood Inundation maps from MODIS (left) and model (right) on the 01 July 2004
Fig. 9 During-flood inundation maps from MODIS (left) and model (right) on the 27th July 2004
Figure 8 shows the before-flood situations. On the left, there was no flood in the
MODIS inundation map on 01 July 2004. Here, the green color shows the non-water
zone, whereas the blue area and the green area with blue shade show the flood water
zone. On the right, there is almost no flood in the model-generated inundation map
on the same day. Here, the blue part shows the flood water zone, and the white part
shows the non-water zone (Fig. 8).
The observed average water level on 01 July 2004 was 5.95 mPWD and the
model-generated water level was 5.75 mPWD at 12 pm at Bhagyakul (SW 93.4L)
gage station.
Figure 9 shows the during-flood situations. On the left is the MODIS inundation
map on 27 July 2004 and on the right is the model-generated inundation map on the
same day.
Here also, the blue area indicates the floodwater zone in the figures. On the left
side of Fig. 9, the MODIS image shows that the green areas with any blue shade on
the floodplain were actually inundated. Areas not inundated are marked by circles
for easy visualization. So, almost all areas were flooded except the areas near the 2D
boundary at the upstream right side and downstream left side of the Padma River.
The observed average water level on 27 July 2004 was 7.13 mPWD and the model-
generated water level was 7.28 mPWD at 12 pm at the Bhagyakul gage station on
the same day.
Some higher areas have been flooded in the model which shows some discrepancy
with the satellite image. This may be due to the initial condition of the model or the
DEM quality. In this study, the DEM has been used with 500 m resolution, so some
areas might have become lower at the time of DEM processing or terrain creation.
Otherwise, the comparison between observed flooding from satellite imagery and
simulated flooding from model results is found to be satisfactory.
3.3 Stabilization Impact Analysis
Under the stabilized condition, the river channel will not change its course and there
will be flood-control embankments on both sides of the river. As both the river banks
will be stabilized and flood-control embankments will be constructed, water will not
enter into the floodplains even in a high flood situation. So, the hydraulic behavior of
the Padma River will change under the recommended stabilized condition. Figures 10
to 12 represent the changes in such behaviors of the Padma River. The water will
remain within the river that was previously used to enter into the floodplains. So, the
river will carry more water under the stabilized condition. As a result, the discharge,
water level, and velocity will increase under the stabilized condition. Figure 10 shows
the change in the discharge of the Padma River at Mawa gage station (SW93.5L)
under the stabilized condition.
In Fig. 10, the blue line represents the model-simulated discharge without stabi-
lization at Mawa (present condition). The orange line represents the discharge at
Mawa under the stabilized condition. A considerable rise in discharge is found after
Fig. 10 Change in discharge 150000

at Mawa due to the
Discharge (m3/s)
stabilization of the Padma 100000

River
50000
Without StabilizaƟon
With StabilizaƟon
0
04-Jun-98 24-Jun-98 14-Jul-98 03-Aug-98 23-Aug-98 12-Sep-98
Fig. 11 Change in water 8.0
Water Level (mPWD)

level at Mawa due to the 7.0
stabilization of the Padma 6.0
River 5.0 Without Stabilizaon
4.0 With Stabilizaon
3.0
Fig. 12 Change in 3.0

maximum average velocity 2.5
magnitude near Mawa due to Velocity (m/s)
2.0
the stabilization of the 1.5
Padma River 1.0 Without StabilizaƟon
0.5 With StabilizaƟon

0.0
4-Jun-98 24-Jun-98 14-Jul-98 3-Aug-98 23-Aug-98 12-Sep-98
the 1st July. On average, 16.28% increase in discharge is found from 1 July to 22
September and 20.04% increase in discharge is found at the peak. The daily average
peak discharge without stabilization is found to be 103,766 m3 /s and with stabilization
124,559 m3 /s on 9 September 1998 from the model simulation.
Figure 11 shows the change in the water level of the Padma River at Mawa
under the stabilized condition. Here, the blue line represents the model-simulated
water level without stabilization at Mawa (present condition), and the orange line
represents the water level at Mawa under the stabilized condition (Fig. 11).
A considerable rise in water level is found after 1 July 1998. On average 6.22%
(0.39 m) increase in water level is found from 1 July to 22 September, and 7.14% (0.51
m) increase in water level is found at the peak. The without-stabilization water level
is 7.15 mPWD, and with-stabilization water level is 7.67 mPWD on 9 September
1998 at Mawa.
Figure 12 shows the change in the maximum average velocity magnitude of the
Padma River near Mawa under the stabilized condition. Here, the blue line represents
the model-simulated maximum average velocity magnitude without stabilization
near Mawa (present condition), and the orange line represents the maximum average
velocity magnitude under the stabilized condition (Fig. 12).
A rise in maximum average velocity magnitude is found after 1 July 1998. On
average, 18.67% increase in maximum average velocity magnitude is found from 1
July to 22 September, and 26.74% increase in maximum average velocity magnitude
is found on 8 September. The peak of the maximum average velocity magnitude is
1.99 m/s for without stabilization, and 2.50 m/s for the stabilization condition on
9 September 1998 near Mawa. A 25.43% increase in maximum average velocity
magnitude is found during the peak, while the maximum change occurs on the
previous day.
4 Discussion
This study has developed an important tool for assessing the future hydraulic behavior
of the Padma River (e.g., changed discharge, water level, velocity) under the proposed
stabilized condition. The study reveals that the stabilization will make the floodplain
of the river flood-free. But the discharge, water level, and velocity magnitude will
increase as demonstrated by the findings of this study. This information can be
valuable for designing the hydraulic structures, such as fixing the heights of the
flood-control embankment, bridges, roads, and others within the area.
Along with resource constraints, this study has some further limitations. Though
the Padma River carries a huge amount of sediment load, the impact of sedimenta-
tion, local scour, and erosion have not been considered by the study. The local rainfall
effect and possible climate change impacts in the future have not been considered.
The resolution of the DEM is 500 m, which is also a limiting factor. These limitations
could have resulted in some deviations of the model results from the measured data.
Calibration and validation results show some overestimation of the water level on
the floodplain and hence the inundated area, which is more likely due to the DEM
resolution and quality. The qualitative comparison of the flood inundation map is
also a limiting factor of this study. Quantitative comparison was not possible as the
processed MODIS images were collected from Roy (2019) which is a secondary
source. But the calibration and validation of water level at Bhagyakul on the Padma
River indicate the high reliability of this model. From the values of the coefficient
of model performance evaluation, it can be said that this 1D-2D-coupled HEC-RAS
model setup can simulate the hydraulic behavior of the Padma River and its flood-
plains undoubtedly well. This mathematical model is able to calculate the hydraulic
parameters at all locations within the model domain for both 1D and 2D areas very
well. So, simulations of different possible conditions in the future using this model
setup can be highly acceptable.
Stabilization intended will have beneficial impacts by preventing flood damage
to agriculture and infrastructure within the area. Under the stabilized condition with
flood embankment, the values of the hydraulic parameters will increase. The peak
discharge of 100 years of return period, like the 1998 flood, will be increased by
20% at the Padma River. The water will not enter into the floodplain under the stabi-
lized condition. It was also found during the field survey that the floodplains are
getting little to no water due to the newly built embankment, especially in Shibchar
area of Madaripur district and Jajira area of Shariatpur district. The river training
embankment for the Padma Multipurpose Bridge is blocking the channels that are
used to connect those floodplains with the Padma River. It has been found from
the interviews and group discussions with the local people, that they are not getting
water from those channels. As a result, they are facing water deficiency for growing
crops. The areas are becoming arid day by day, and the local environment is changing
rapidly. Local people anticipate that these channels will be lost in the future. The
people living in the floodplains of the Padma River are largely dependent on agri-
culture and fishery-based livelihoods. They are already experiencing the negative
impacts of the embankment.
So, before the construction of the flood control and river stabilization embank-
ment, the environmental impact on the entire region should be assessed, as several
negative impacts may occur in the future. Floodplain aquatic habitats can be degraded
due to reduced flooded areas and loss of hydrologic connectivity. This will adversely
affect the floodplain-dependent open-water fish migration as well as wetland biodi-
versity. These impacts can in turn adversely affect the nutrition, health, and economic
status of poor people in the area. Drainage congestion problems might occur as water
may be entrapped in the area due to the construction of a new embankment. There is
a need to develop a comprehensive river stabilization plan to identify potential stabi-
lization solutions with minimal impacts on the river and floodplain environment. The
off-takes of the distributaries can be stabilized in such a way that it allows stable dry
season flows to pass and normal beneficial flooding to occur during the monsoon
season, while restricting high damaging floods through regulations.
5 Conclusion
The Government of Bangladesh has already undertaken many initiatives for better
water resource management and disaster risk reduction in the country. Some of these
initiatives include the policy regarding stabilization and channelization of the Padma
River. The stabilization plans include rehabilitation and construction of flood-control
embankments, which can change the hydraulic behavior of the river. The stabiliza-
tion impacts can be simulated and analyzed through a hydrodynamic model like
HEC-RAS. In a country like Bangladesh, where data are very limited and resources
are scarce, such modeling approaches can give reasonable predictions of required
scenarios.
In this study, an HEC-RAS-coupled 1D-2D model of the Padma River system has
been developed for evaluating the stabilization impacts on the hydraulic behavior of
the river. The rivers have been modeled as a 1D flow area, and the floodplains have
been modeled as a 2D flow area. Model calibration has been performed for the years
1998, and validation for the year 2004, where Manning’s roughness coefficient ‘n’
was the main calibration and validation parameter. Further calibration and validation
at the floodplain have been done using the observed highest flood water level, and the
flood inundation area with the MODIS images, respectively. The results have shown
a good performance of the model.
After calibration and validation, the model has been run according to the stabi-
lization plan. The floodwater cannot enter into the floodplain and remains within
the cross-sectional area under the stabilized condition. This results in an increase in
discharge, water level, and velocity in the river. It is found that the peak discharge of
the 1998 flood can increase by 20% with an average increase of 16% at Mawa on the
Padma River. On average, 0.39 m increase in water level has been found with 0.51
m increase in the peak water level at Mawa. The average increase in the maximum
average velocity has been found to be 19%, with the increase of 25% in the peak
water level for the Padma.
The findings from this study will provide valuable information regarding the future
stabilization impacts on river hydraulics in the Padma River and its floodplain. It is
expected that the outcomes of this study will be useful for the authorities responsible
for flood management and disaster risk reduction in the study area.
Acknowledgements The authors would like to thank all the survey respondents and participants
of the group discussions, for providing valuable information. Special thanks go to Mr. Santosh
Karmaker, a former student of the Department of Civil Engineering, BUET, Mr. Shahriar Shafayet
Hossain, a former student of IWFM, BUET, and the IWFM survey team for accompanying the
authors during the field survey. The authors are also grateful to the faculties and laboratory technician
of IWFM, BUET for their assistance. Finally, the valuable comments and suggestions provided by
the anonymous reviewer are gratefully acknowledged.
References
Baird, Drew C., Lisa Fotherby, Cassie C. Klumpp, and S. Michael Sculock. 2015. Bank Stabilization
Design Guidelines. Denver: U.S. Department of the Interior, Bureau of Reclamation Technical
Service Center.
Biswas, Subir, M. Shahjahan Mondal, and Md. Rashedul Islam. 2020. Simulating Flood Risk Due
to Climate Change in the Padma River System Using the 2D HEC-RAS Model. 5th International
Conference on Civil Engineering for Sustainable Development (ICCESD), 7–9 February 2020,
Khulna.
Brunner, Gary W. 2014. Combined 1D and 2D Modeling with HEC-RAS. Davis, CA: US Army
Corps of Engineers, Hydrologic Engineering Center.
Brunner, Gary W. 2016. HEC-RAS River Analysis System, 2D Modeling User’s Manual. Davis,
CA: US Army Corps of Engineers, Hydrologic Engineering Center.
BWDB (Bangladesh Water Development Board). 2011. Rivers of Bangladesh. Dhaka, Bangladesh.
BWDB. 2014. Bangladesh: Flood and Riverbank Erosion Risk Management Investment Program.
Dhaka, Bangladesh.
BWDB. 2019. Sixty Years of Water Resource Management & Development in Bangladesh. Dhaka,
Bangladesh.
FFWC (Flood Forecasting and Warning Centre). 2018. Annual Flood Report 2018. Flood Fore-
casting & Warning Centre, Processing & Flood Forecasting Circle. Dhaka: Bangladesh Water
Development Board.
GED (General Economics Division). 2017. Bangladesh Delta Plan 2100. Bangladesh Planning
Commission, Government of the People’s Republic of Bangladesh. Dhaka, Bangladesh.
Hossain, A. N. H. Akhtar. 2003. Integrated Flood Management Case Study Bangladesh: Flood
management. Technical report. The Associated Programme on Flood Management, World
Meteorological Organization.
Islam, M. N. Nabiul. 2011. Impacts of Urban Floods from Micro-Macro Level Perspectives: A Case
Study of Bangladesh. Chisinau: Lambert Academic Publishing.
IWFM. 2018. Hydrology and Morphology Study Work Under Detail Feasibility Study for
Bangabandhu Sheikh Mujib International Airport Project. Bureau of Research: Testing and
Consultation, Bangladesh University of Engineering and Technology, Dhaka, Bangladesh.
McLean, D.G., Jose A. Vasquez, Knut Oberhagemann, and Maminul H. Sarker. 2012. Padma
River Morphodynamics Near Padma Bridge. River Flow 2012—Proceedings of the International
Conference on Fluvial Hydraulics 1: 741–747.
Mondal, M. Shahjahan, A.K.M. Saiful Islam, Anisul Haque, Md. Rashedul Islam, Subir Biswas,
and Khaled Mohammed. 2018. Assessing High-End Climate Change Impacts on Floods in Major
Rivers of Bangladesh Using Multi-Model Simulations. Global Science and Technology Journal
6 (2): 1–14.
Moriasi, Daniel N., Jeffrey G. Arnold, Mi-chael W. Van Liew, Ronald L. Bingner, R. Daren Harmel,
and Tamie L. Veith. 2007. Model Evaluation Guidelines for Systematic Quantification of Accu-
racy in Watershed Simulations. Transactions of the ASABE 50 (3): 885–900. https://doi.org/10.
13031/2013.23153.
NHC (Northwest Hydraulic Consultants). 2013. Bangladesh: Main River Flood and Bank Erosion
Risk Management Program: Final Report, Main Report. Dhaka.
Ninno, Carlo del, Paul A. Dorosh, Lisa C. Smith, and Dilip K. Roy. 2001. The 1998 Floods in
Bangladesh: Disaster Impacts, Household Coping Strategies, and Response. Research report
No.: 122. Washington D.C.: International Food Policy Research Institute.
Nishat, Ummay. 2017. Flood Inundation Mapping of Jamuna River Floodplain Using HEC-RAS
2D Model (B.Sc. Thesis). Dhaka: Department of Water Resources Engineering, Bangladesh
University of Engineering and Technology.
Ponce, V.M. 1990. Generalized Diffusion Wave Equation with Inertial Effects. Water Resources
Research 26 (5): 1099–1101.
Quiroga, V., S. Moya, K. Udo. Kure, and A. Mano. 2016. Application of 2D Numerical Simulation
for the Analysis of the February 2014 Bolivian Amazonia Flood: Application of the New HEC-
RAS Version 5. RIBAGUA-Revista Iberoamericana Del Agua 3 (1): 25–33. https://doi.org/10.
1016/j.riba.2015.12.001.
Roy, Binata. 2019. A Study on Fluvial Flood Hazard and Risk Assessment of Arial Khan River
Floodplain Under Future Climate Change Scenarios (M.Sc. Thesis). Dhaka: Department of
Water Resources Engineering, Bangladesh University of Engineering and Technology.
Tazin, Tasmia. 2018. Flood Hazard Mapping of Dharala Floodplain using HEC-RAS 1D/2D
Coupled Model (M.Sc. Thesis). Dhaka: Department of Water Resources Engineering, Bangladesh
University of Engineering and Technology.
Water and Livelihood Security
A Sustainability Index for Assessing
Village Tank Cascade Systems (VTCs)
in Sri Lanka
E. M. G. P. Hemachandra, N. D. K. Dayawansa,
and Ranjith Premalal De Silva
Abstract Traditional Village Tank Cascade systems (VTCs) are a connected series
of tanks where the excess water in upstream tanks is collected and reused by the
downstream tanks. These are highly sustainable irrigation systems in the Dry zone
landscape of Sri Lanka from ancient times. At present, these systems are under threat
of deterioration due to continuous human interventions. Even though a number of
studies have identified different parameters to assess VTCs, none of them have
addressed the social, environmental and economic sustainability within a single
assessment. Since the sustainability of VTCs should achieve the equilibrium of its
ecological, hydrological, agricultural and social subsystems and cultural norms, it is
important to incorporate all these factors in the sustainability assessment. Consid-
ering this background, the objective of this study was to develop an index to assess
the sustainability of VTCs in an integrated manner. The index was developed with 8
physical-environmental parameters and 13 socioeconomic parameters. The relative
importance of each parameter was identified through a Multi-Criteria Analysis. The
developed index can be used in assessing the sustainability of any of the VTCs as
well as the individual tanks in a 0 to 100 scale, which can be used for comparing and
prioritizing the cascades for rehabilitation and development activities.
Keywords Village Tank Cascade Systems (VTCs) · Sustainability · Index

Development · Parameters
1 Introduction
The undulating landform in the dry zone of Sri Lanka has allowed ancient kings
to construct village tanks/tank cascade systems (VTCs) by obstructing the natural
E. M. G. P. Hemachandra (B)
Postgraduate Institute of Agriculture, University of Peradeniya, Peradeniya, Sri Lanka
N. D. K. Dayawansa · R. P. De Silva
Department of Agricultural Engineering, Faculty of Agriculture, University of Peradeniya,
Peradeniya, Sri Lanka
300 E. M. G. P. Hemachandra et al.
drainage system for the purpose of irrigation and other rural water requirements.
According to the definition of Madduma Bandara (1985) “Cascade System is a
connected series of tanks organized within a meso-catchment of the dry zone
landscape, for storing, conveying, and utilizing water from an ephemeral rivulet.”
The upstream tanks in the cascade system collect water from rainfall and runoff
while the downstream tanks receive water due to the drainage, seepage, surface runoff
and direct rainfall to the tank. Therefore, the excess water of an upstream tank used
by its command area is drained to the downstream tank and used for irrigating its
command area (Panabokke et al., 2001). A representative diagram of a village tank
cascade system is presented in Fig. 1.
The communities living in these areas have very close interactions with the VTCs
since most of their livelihood activities are closely associated with village tanks
and their ecosystems. Considering the ecological components and the sustainable
land use system associated with the VTCs, they were once considered as highly
sustainable agro-ecosystems. As a result, communities reaped multiple benefits from
them including efficient rainwater storage to be utilized during dry season, flood
control, provision of water for agriculture, domestic needs, livestock, maintenance
of a high groundwater level, facilitation of growth of reeds, etc. Thus, the VTCs
are considered as multi-purpose systems (Panabokke et al., 2001). Hence, based
on specific functions and components of VTCs, Dharmasena (2020) has defined
the cascade system as an ecosystem in the dry zone landscape of Sri Lanka which
sustainably fulfills basic needs to humans, plants and animals using water, soil, air
and vegetation.
Even though these systems were sustainable in the past, they are gradually dete-
riorating due to intense anthropogenic activities (Dharmasena, 2010). The level of
deterioration is varying from cascade to cascade hence prioritization is important
in restoration/rehabilitation activities. Since the sustainability of the cascade system
Fig. 1 Representative
diagram of a village tank
cascade system (Source
Bandara et al., 2008)
A Sustainability Index … 301
includes the equilibrium of its subsystems such as ecological, hydrological, agricul-

tural, social systems and cultural norms (Marambe et al., 2012), incorporation of all
these factors is needed to assess the sustainability.
A number of parameters have been identified in the past studies to assess the
sustainability of VTCs (Sakthivadivel et al., 1994; Dharmasena and Goodwill, 1999;
Dharmasena, 2009; Perera, 2016). In general, social, environmental and economic
factors collectively influence the sustainability of tanks/VTCs, and hence, it is diffi-
cult to consider only one group of factors in sustainability assessment. Integration of
these different parameters into a single assessment strategy has not been addressed
in the past studies. An index which can incorporate all possible parameters which
are influential on the sustainability of VTC/individual village tanks will be helpful
to identify and compare the status of these systems. These assessments will facilitate
the managers to focus on most susceptible systems to introduce remedial measures.
In addition, this type of sustainability assessment method will support continuous
monitoring of these systems to obtain optimum services from them. Accordingly,
planning and management of the VTCs should be done with a broad understanding
of principles and mechanisms of the cascade system functions (Geekiyanage and
Pushpakumara, 2013).
2 Methodology—Development of a Sustainability Index

to Assess VTCs
The VTCs consist of a variety of subsystems, which include the ecological system
of forests, aquatic habitats and common areas, the water management system of
structures and the system of water distribution, the agricultural system, the social
system with well-established formal and informal institutional setup and the cultural
and spiritual norms which are helpful for simple and conflict free life (Marambe et al.,
2012). Therefore, the sustainability of VTCs is a measure of how these subsystems
utilize the resources to provide services while maintaining a sustainable ecosystem
balance.
2.1 Identification of Variables/Parameters of Sustainability

of VTCs
Twenty-one parameters of sustainability of VTCs were considered by using existing

parameters, improving the parameters and developing new parameters. Four of them
were taken as it is from the existing literature and seven parameters were newly
developed by improving the existing parameters. Additionally, ten new parameters
that have not been considered in past studies were developed to include the essential
factors which influence the sustainability of the VTCs. All the parameters used in this
Table 1 Parameters of sustainability index of VTCs

No Existing parameters New parameters developed Newly developed parameters
by improving existing
parameters
1 Ratio of Catchment area: Live storage of the Presence of aquatic plants in
Water surface area of tank tank/Total irrigation village tanks (% coverage)
(Sakthivadivel et al., 1994) command area
(Dharmasena, 2009)
2 Frequency of spilling the Existence of ecological Existence and Performances
tanks within a specific period components (Sangrama, of farmer organizations
of time (per year) 2018)
(Sakthivadivel et al., 1994)
3 Agro-well density Percentage of forest cover Healthy agricultural
(Dharmasena and Goodwill, in the catchment (Perera, practices—Organic fertilizer
1999) 2016) usage
4 Cropping intensity (paddy) Presence of risky land uses Percentage of farmers who
(Sakthivadivel et al., 1994; in the catchment practice in water
Dharmasena, 2009) (Samarasinghe and conservation strategies
Dayawansa, 2011)
5 Population density in the Percentage of young farmers
cascade system (Jayarathne involved in agriculture
et al., 2010)
6 Severity of human-elephant People’s awareness of
conflict (Perera, 2016) cascade system functions
7 Percentage of Samurdhi Women’s participation in
beneficiaries (Jayarathne cascade activities
et al., 2010)
8 Multiple water uses/users
9 Average paddy yield in Maha
(Major cultivation) season
10 Number of cascade related
projects implemented during
last 10 years
study are presented in Table 1. A detailed description of the parameters is presented

under “Development of sustainability Index.”
2.2 Development of the Sustainability Index
A conceptual framework of the sustainability index was developed incorporating all

the contributing parameters (Fig. 2). The parameters were then categorized into two
major groups of physical-environmental parameters and socioeconomic parameters,
based on the applicability of remedial and rehabilitation measures of VTCs after the
sustainability assessment. It would be difficult to address physical and environmental
Fig. 2 Conceptual framework of developing the sustainability Index
parameters as they are high costly and will require political interventions and the
approvals of many institutions and authorities. However, with the help of individuals
or community-based organizations in a given cascade system, there is more flexibility
when it comes to addressing social and economic parameters.
2.2.1 Calculation of Parameters and Normalization Scales
Different parameters are measured in different units and are in different scales. When
developing an index, it is important to bring them to a common scale to identify
their contribution to VTC sustainability. Therefore, the actual parameter values were
converted into a 0–100 linear scale as stated below.
• Physical and Environmental Parameters
I. Ratio of Catchment area: Water surface area of tank
This parameter is looking for the hydrologic potential of the individual tanks or
the tank cascade system. According to Sakthivadivel et al., (1994) a ratio of 7.5
is required to have an adequate hydrological potential. It implies that the area of
the catchment should be 7.5 times as large as the water surface area of the tank
which is a measure of the capacity of a small tank. Accordingly, a ratio larger than
7.5 would have high hydrologic potential and less than that value would have low
hydrologic potential (Sakthivadivel et al., 1994). Assuming that with the reduction of
the ratio, the sustainability of the cascade is reducing, the normalized score (linear)
was developed according to Table 2.
II. Ratio of live storage of the tank: Total command area
Based on the major uses of a reservoir, its capacity is divided into three components:
live storage/active storage, dead storage and flood storage. The live storage is used to
supply water to the downstream, hydropower generation and recreational purposes.
Table 2 Score of “Ratio of Catchment area: Water surface area of tank” and “Ratio of Live storage
of tank (ha.m): Command area (ha)”
Category Ratio of Catchment area: Water Ratio of Live storage of tank (ha.m): Score
surface area of tank Command area (ha)
i <1 < = 0.25 0
ii 1– < 2.5 0.25–0.5 25
iii 2.5– < 5 0.5–0.75 50
iv 5– < 7.5 0.75–1 75
v > = 7.5 >=1 100
The dead storage is used for sediment deposition, while the flood storage is used
for controlling floods and its adverse impacts to the downstream (Loucks and Van
Beek, 2017). Therefore, only the live storage is available for irrigation water supply
in village tanks. When the command area increases compared to the live storage of
the tank, then there is high pressure on irrigation water requirement. This can often
result in water scarcity. On-farm water requirement of 3-month paddy variety in Maha
season under Reddish Brown Earth (RBE) moderately drained soil is 1057 mm, and
it is 948 mm under Low Humic Gley soil (LHG) (https://doa.gov.lk/). Accordingly,
considering the adequacy of water for cultivation during Maha season, the score
value scale was developed for the parameters (Table 2).
III. Frequency of spilling the tanks within a specific period of time (per year)
People in village tank cascade systems highly depend on the availability of tank water
for agriculture and other livelihood activities. Therefore, people expect the tank to
be filled during the rainy season. If there is adequate rainfall, farmers can cultivate
during Maha season mainly with rainwater, and the excess water can be stored in
the tanks for cultivating during Yala (relatively dry minor cultivation season) season.
Assuming that tanks should spill minimum once per year most probably during
North-East monsoon rains, a score of 100 was given to the tanks which spill once
or more than once per year. A score of 0 was given when the tanks do not spill at
least once a year. The timing and duration of spilling differ with the location of the
tank in the cascade system. The downstream tanks spill for a long time compared
to upstream tanks since they receive water from a large catchment area above them.
However, the spilling duration was not considered as a parameter in the sustainability
index.
IV. Existence of ecological components
Essential ecological components of VTCs such as Interceptor, Tree belt, Water hole,
Upstream side ridges and Tank bund provide beneficial effects to the tank cascade
system, thus important for their sustainability as well (Dharmasena, 2010). Accord-
ingly, the existence of these ecological components in individual tanks is assessed
by this parameter. The given normalization scores for the existence of ecological
parameters were “0” for “Not existing/ unsatisfied,” “50” for “Existing but moder-
ately satisfied” and “100” for “Existing and satisfied.” The scores are given to all
five ecological components and their average value is considered as the parameter
value in the index. When the index is applied to an entire tank cascade system,
the average value for the ecological components of all individual tanks in the same
cascade system is taken.
V. Agro-well density in the catchment
Agro-wells are constructed to be used as a source of supplementary irrigation in most
of the catchments in dry and intermediate regions in Sri Lanka (Pelpitiya et al., 2016).
However, overabstraction of water in the upper catchment area can influence water
availability in the whole catchment. When this is happening in cascade systems, it
can adversely influence the sustainability of the system. To maintain the sustain-
ability, the recommended agro-well density is 7–8 wells/km2 (7–8 wells per 100 ha)
(Dharmasena and Goodwill, 1999). The score scale was developed according to this
understanding (Table 3).
VI. Percentage of Forest cover in the catchment
The catchment land use quality is highly important to achieve a better hydrologic
potential. The forests play a major role in hydrology and can increase the infiltration
rate, and they can increase the water availability within the catchment (Ekhuemelo
et al., 2016; Bonnesoeur et al., 2019). Therefore, this parameter is emphasizing the
fact that, with the increment of the forest cover in the catchment, the sustainability
of the cascade will increase. Accordingly, the scores were identified (Table 3). When
the index is applied for a total cascade, the forest cover of the whole catchment of
the cascade system is taken.
VII. Presence of risky land uses in the catchment
Since the cascade system acts as a single ecosystem, a risky human intervention in
one location of the catchment can influence the whole biotic and abiotic components
in different ways. This parameter looks for the existence of risky land uses such as
large-scale constructions, industrial sites, large-scale deforestations. The parameter
was developed to qualitatively state whether the risk is low, moderate, high or very
high (Table 3).
Table 3 Score for “Agro-well density in the catchment,” “Percentage of forest cover in the
catchment” and “Presence of risky land uses in the catchment”
Category Agro-well density Percentage forest cover Presence of risky land Score
(wells per km2 ) in the catchment uses in the catchment
i > = 50 0% Very high risk 0
ii 30–50 1%– < 25% High risk 25
iii 15–30 25%– < 50% Moderate risk 50
iv 8–15 50%– < 75% Low risk 75
v <=8 > = 75% Not existing 100
VIII. Presence of aquatic plants in tanks (% coverage)

Aquatic plant distribution in tanks can be taken as both positive and negative for
the sustainability of the system. In ancient times, people largely benefited from
aquatic plants such as lotus (Nelumbo nucifera), Water lily (Nymphaea) and Kekatiya
(Aponogeton crispus). They could also get economic benefits (flowers can be sold for
religious offerings) and food security (some plant parts can be used as food items).
However, there are a limited number of people who gain these benefits in the modern
society.
In addition, some other plants such as Water hyacinth (Eichhornia crassipes) can
trap sediments and fill the tanks to reduce the capacity, block the irrigation canals
and can also act as weeds in agricultural fields when transported with irrigation
water. When the water surface is covered with aquatic plants, water can be lost
through evapotranspiration. Spreading of aquatic weeds in water bodies can also
cause economic and ecological losses to the system (Wijesundara, 2010). Since the
negative impacts associated with aquatic plants distribution are severe and it can
affect a greater number of people and the water circulation within the overall system,
the presence of plants was considered as negative for the sustainability of VTCs.
Table 4 presents the score given to this parameter.
• Socioeconomic parameters
I. Existence and performances of farmer organizations

In small irrigation systems, the basic functions of water and cultivation management
are done through the farmer’s organizations under the guidance of the Department
of Agrarian Development. Therefore, it is important to have farmer’s organizations
for the sustainability of VTCs.
Additionally, the performances of the farmer’s organizations are very important
for better management of water and other resources in the cascade system and the
maintenance of the tanks which subsequently uplift the livelihood of the farmers.
The performance of the farmer’s organizations is a function of farmer participation
in meetings, balanced gender representation in the executive committees, major func-
tions/activities related to cascade systems and funds/savings. Accordingly, criteria
such as existence of farmer organization, farmer participation in meetings, gender
representation in the executive committee, major functions and activities, water
management related disputes and available funds were considered to assess this
parameter and the final mark itself was considered as the index score.
Table 4 Score of “Presence of aquatic plants in tanks”

Category Percentage coverage of aquatic plants in tanks Score
i > = 75% 0
ii 50%– < 75% 25
iii 25%– < 50% 50
iv < 25% 100
Table 5 Score of “Organic fertilizer usage” and “Percentage of young farmers involve in
agriculture” and “Multiple water uses/users”
Category Percentage of farmers Percentage of young Multiple water Score
who use organic farmers involve in uses/users:- Number of
fertilizer agriculture benefits from the
cascade system/ tank
i 0% 0% 0 0
ii 1%– < 10% 1%– < 10% 2– < 4 20
iii 10%– < 20% 10%– < 20% 4– < 6 40
iv 20%– < 30% 20%– < 30% 6– < 8 60
v 30%– < 50% 30%– < 50% 8– < 10 80
vi > = 50% > = 50% > 10 100
II. Healthy Agricultural practices—Organic Fertilizer Usage

There is reuse of water in the tank cascade systems. The excess water in the upstream
paddy fields is drained to the downstream tanks and that water is utilized in its
command area. Therefore, the agro-chemical usage in the upstream area has influ-
ences the water quality of the whole cascade system. If higher number of organic
farmers are in the cascade, there is less chance of polluting water from heavy doses
of inorganic fertilizers. Since complete organic farming is very rare in these cascade
systems, farmers who attempt to incorporate organic fertilizers for their farming were
considered to be promoting a healthy agricultural practice. Therefore, the following
score was developed to evaluate the impact of organic fertilizer usage (Table 5).
III. Percentage of farmers who practice water conservation strategies
The most upstream tanks which do not connect through a drainage canal in the cascade
system are highly dependent on rainwater. Farmers practice some water conservation
measures such as bund maintenance, proper drainage of water, timely cultivation,
mulching, incorporation of organic matter with soil. Mostly in Yala season, farmers
have to conserve water and Sri Lankan Dry Zone farmers practice “Bethma” as
a strategy to cultivate under water scarce situation (Somarathna and Kono, 2005).
Under the “Bethma” system, a part of the command area is temporarily distributed
among all the farmers in the command area based on the available water in the tank
during dry seasons (Oka et al., 2015). Percentage of farmers who practice water
conservation strategies itself was considered as the index score.
IV. Percentage of young farmers involved in agriculture
Reluctance of the young generation to take up farming as the occupation is one of the
common problems in many parts of Sri Lanka (Somaratne, 2002). When there are no
adequate people for agricultural activities, eventually the maintenance and manage-
ment of these small tank systems will face problems. Hence, this parameter was
considered as important with respect to the sustainability of the tanks/tank cascades.
The farmers below 40 years of age were considered as young farmers (Zondag et al.,
2015). The scale used to provide scores to this parameter is presented in Table 5.
V. Multiple water uses/users

Livelihood of the ancient people was highly related to the village tank cascade func-
tions. The multiple benefits of VTCs are water for irrigation, bathing, washing and
drinking, sand mining, collecting firewood, harvesting sedges, collecting bee honey,
maintaining groundwater table, harvesting lotus/water lily, harvesting timber, water
for livestock, etc. This parameter identified the multi-purpose nature of the tanks
in the cascade which is related to multiple water uses/users. Table 5 shows the
normalization score of this parameter.
VI. Population density
Increment of population leads to overutilization of natural resources, thereby
increasing environmental pollution, land encroachment and social conflicts. To avoid
these problems in ancient VTCs, the village margins have been demarcated. There-
fore, people were not allowed to encroach common lands and disturb the land use
plan of the tank cascade system. This parameter considers the impact of popula-
tion density on the sustainability of the tank cascade system. Table 6 presents the
normalization score used for this parameter.
VII. People’s awareness of cascade system functions
Traditional communities who have evolved around the village tanks used to appre-
ciate the functions of the cascade systems since they had a good understanding of
them. With time, this close relationship between tanks and the people was broken
and the traditional knowledgebase also faded away. Lack of awareness about the
functions of the cascade system and its components is also a parameter which is
responsible for the sustainability of these systems. Knowledge about the VTCs,
water utilization by different stakeholders in that area including downstream people,
wild animals and overall environment are key factors with respect to the awareness.
Understanding and awareness of the functionality of the entire cascade system lead to
overcome the weaknesses and challenges for sustainability (Marambe et al., 2012).
Accordingly, an assessment score was developed based on the awareness of defining
the cascade system, identifying the different stakeholders and understanding of the
multiple benefits of the cascades system.
VIII. Women engagement in cascade activities
According to third Dublin principle, 1992 “Women play a central part in the provi-
sion, management and safeguarding of water” (https://www.gwp.org/). It shows that
Table 6 Score of
Category Population density (person per sq. km) Score
“Population density”
i > = 400 0
ii 300– < 400 25
iii 200– < 300 50
iv 100– < 200 75
v < 100 100
women play a key role in water management related activities. However, women’s
involvement in the farmer organizations activities, water management related deci-
sion making and problem solving are not substantial in many water management
systems including cascades (https://www.gwp.org/). However, there is a tendency
for females to protect water and natural resources since they are more concerned
about the well-being of their families. Hence, this parameter is qualitatively looking
for the level of women’s engagements for the cascade activities and is considered as
positively influencing the sustainability of the system. Accordingly, the scores for
“Women engagement in cascade activities were given as “0” for “Not satisfied,” “50”
for “Moderately satisfied,” “75” for “Satisfied” and “100” for “Highly satisfied.”
IX. Severity of human-elephant conflict
Human-Elephant Conflict (HEC) is one of the national problems in Sri Lanka at
present and would become worse in the future (Santiapillai et al., 2010). Mostly the
elephant’s damages to the cultivations are high in Dry season (Campos-Arceiz et al.,
2009). Cultivations are less in dry season and elephants tend to find the few cultivated
areas to fulfill their food requirements. Farmers are reluctant to cultivate when the
HEC is high. Within a cascade system if the cultivations are damaged or abandoned
due to HEC, it will have an adverse impact on the economy. Subsequently, it also has
an impact on water and other resource management in the cascade systems. Some
tanks can be completely abandoned as a result of the elephant threat (Anuradha et al.,
2019). Additionally, HEC can be a huge threat to other domestic and livelihood activ-
ities. This parameter identifies the severity of HEC with respect to the sustainability
of the tanks. Table 7 shows the normalization scale of HEC in the index.
X. Percentage of Samurdhi beneficiaries
Samurdhi is a program launched in 1995 by the Sri Lankan government to reduce
poverty by providing food stamp and other financial supports for selected households
based on their income levels. Therefore, the number of Samurdhi beneficiaries is an
indirect parameter to measure existing poverty in the area. Hence, this parameter
reflects the economic status of the area. When the poverty is high, people tend to
exploit natural resources to fulfill their daily needs, thereby making natural systems
Table 7 Score of “Severity of HEC,” “Percentage of Samurdhi beneficiaries” and “Average paddy
yield in Maha season”
Category Severity of HEC: Percentage of Samurdhi Average paddy yield in Score
Percentage of people beneficiaries Maha season (kg per ha)
affected by HEC
i > 75% > 75% < 3000 0
ii 50%– < 75% 50%– < 75% 3000– < 3500 25
iii 25%– < 50% 25%– < 50% 3500– < 4000 50
iv 0%– < 25% 1%– < 25% 4000– < 4500 75
v 0% 0% > = 4500 100
more vulnerable. Following scale shows the normalization of percentage of Samurdhi

beneficiaries (Table 7).
XI. Average paddy yield in Maha season
People are mostly cultivating paddy in Maha season under minor irrigation systems
due to adequate presence of water with North East monsoon (Sakthivadivel, et al.,
1996). Within a tank cascade system, the water management practices, water quality,
the agronomic practices, etc. will influence the crop yield. Accordingly, the Maha
season paddy yield was taken as a parameter of the sustainability of VTCs. Average
paddy yield in Maha season in Sri Lanka is 4364 kg per ha (http://www.statistics.
gov.lk). Accordingly, following scale was used to bring the paddy yield to measure
sustainability (Table 7).
XII. Cropping intensity (paddy)
Cropping intensity can be defined as the ratio of area cultivated within a year to total
command area. If both Yala and Maha are cultivated, the cropping intensity will
become 2 which is generally the maximum value of cropping intensity for paddy.
Hence the value varies in between 0 and 2. The cropping intensity depends on the
water availability and water management of the cascade system. The index score was
developed by using Eq. 1.
Cr opping I ntensit y
I ndexscor e = ×100 (1)
2
XIII. Number of cascade related projects implemented during last 10 years
The government and non-government agencies are implementing different projects

to rehabilitate the village tank cascade systems. Those projects are implemented
for different purposes such as renovation and rehabilitation of tanks, developing
livelihood of the people in the area, improving natural environment. Since the number
of projects which has been implemented would directly influence the sustainability of
the VTCs, it was considered for developing the normalization score (Implementing
none of the projects = 0, implementing 1–5 projects = 50 and implementing more
than 5 projects = 100).
2.2.2 Calculation of Weights for the Parameters—Analytic Hierarchy

Process
All the above-mentioned parameters are not contributing to the sustainability of the
tanks/cascade systems in an equal scale. Hence, it is important to identify their relative
contribution to the sustainability. An Analytic Hierarchy Process was used in identi-
fying relative importance of these parameters by calculating weightages (Ishizaka and
Labib, 2011; Srdjevic and Jandric, 2010). The terms of the fundamental Saaty’s scale
(Srdjevic and Jandric, 2010) were used to develop the pair-wise comparison matrix.
Accordingly, expert judgments were used to develop weightages for the parame-
ters. However, relative importance of these parameters can change under individual
cascade systems.
2.2.3 Aggregating the Parameters to Develop Index of Sustainability
The physical and environmental sustainability and socioeconomic sustainability were

assessed separately by considering 50% contribution from each. The final sustain-
ability index value was obtained, which is in between the range of 0–100. The index
value indicates whether the tank cascade system/the individual tank is in very low
sustainability (0–25)/ low sustainability (25–50)/ moderate sustainability (50–75) or
high sustainability (75–100).
The developed sustainability index is presented by Eq. 2. It presents all the parameters
with their weightages indicating the relative importance.
Sustainabilit y I ndex o f V T Cs
⎛⎛ ⎞ ⎞
(0.34 × Ratio o f catchment ar ea : W ater sur f ace ar ea o f tank)
⎜⎜ +(0.17 × Ratio o f Live storage o f the tank : Command ar ea :) ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ × within ⎟ ⎟
⎜⎜ 0.07 Fr equency o f spilling the tanks a speci f ic ⎟ ⎟
⎜⎜ + ⎟ ⎟
⎜⎜ period o f time( per year ) ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
=⎜⎜ +(0.11 × Agr o − well densit y in the catchment) ⎟×50%⎟
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ +(0.09 × E xistence o f ecological components) ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ +(0.16 × Per centage o f For est cover in the catchment) ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
⎝⎝ +(0.03 × Pr esence o f risky land uses in the catchment) ⎠ ⎠
+(0.02 × Pr esence o f aquatic plants in tanks)
⎛⎛ ⎞ ⎞
(0.21 × E xistance and per f or mances o f f ar mer organi zations)
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ + 0.16 × H ealthy agricultural practices − Organic f er tili zer ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ usage ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ 0.12 × Per centage o f f ar mer s who practice ⎟ ⎟
⎜⎜ + ⎟ ⎟
⎜⎜ water conser vation stategies ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ +(0.09 × Percentage o f young f ar mers in f armer organi zation) ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ +(0.1 × Population densit y) ⎟ ⎟
⎜⎜
⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ + 0.07 × People s awareness of cascade system functions ⎟ ⎟
+⎜ ⎜ ⎟×50%⎟
⎜⎜ +(0.06 × W omen engagement in cascade activities) ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ +(0.04 × Multi ple water uses and user s) ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ +(0.04 × Severit y o f human − elephant con f lict) ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ +(0.04 × Per centage o f Samur dhi bene f iciaries) ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ +(0.02 × Average paddy yield in Maha season) ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ +(0.03 × Cr opping intensit y( paddy)) ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ 0.02 × N umber o f cascade r elated pr ojects which has ⎟ ⎟
⎝⎝ ⎠ ⎠
+
implemented during last 10 year s
(2)
Figure 3 presents the Microsoft Excel interface of the developed index, applica-
tion of an example data set and the presentation of results. Accordingly, it visualizes
the input value (given example data set), its normalized scale, criteria weightage of
each parameter, separately weighted assessment rank of each parameter and sustain-
ability scores. According to the given example cascade system, overall sustainably
was at the “moderate sustainable level” (sustainability score = 64). It further inter-
prets that the physical-environmental sustainability was at highly sustainable level
(sustainability score = 81) and socioeconomic sustainability was at “low sustain-
ability level” (sustainability score = 47). Accordingly, it leads to the consideration
of factors that can be changed. However, in general, physical and environmental
factors are difficult to change without any financial/government/political assistance.
Hence, practical and low cost approaches are needed to improve the socio-economic
sustainability of the system.
Normalized scale of the individual parameters (Fig. 3) themselves presents the
rehabilitation needs of the cascade system/the individual tank. According to this
particular example, the parameters showing low scores such as existence of ecological
components (score = 40), distribution of aquatic plants (score = 0), existence and
performance of farmer organizations (score = 30), healthy agricultural practices
(score = 40), percentage of farmers who practice water conservation strategies (score
20) and percentage of young farmers involved in agriculture (score = 40) can be
considered in improving the sustainability.
The sustainability index of VTCs can be applied to any VTCs and for indi-
vidual village tanks for quantifying their sustainability, and accordingly compar-
isons between tank cascades/tanks can be done. This can be helpful in selecting
Sustainability Assessment Matrix for Village Tank Cascade Systems

Weighted
Input Normalized Criteria
No Parameter Assessment Sustainability of the tank
value scale weight
rank
1 RaƟo of Catchment area: Water surface area of tank 6 75 0.34 25.65
2 RaƟo of live storage of the tank: total command area 1.5 100 0.17 16.72
Frequency of spilling of the tanks within a specific period of Ɵme
1 100 0.07 7.34
3 (per year)
4 Existance of ecological components
5.1 Interceptor (KaƩakaduwa) 50
Physical and 5.2 Tree belt (Gasgommana) 50
40 0.09 3.75
environmental 5.3 Water hole (Godawala) 0 81
sustainability 5.4 Upstream side ridges (IsweƟ) 0
5.5 Tank bund (Wew kandiya) 100
5 Agro well density in the catchment (number per km2) 9 75 0.11 8.08
6 Percentage of Forest cover in the catchment 80 100 0.16 16.03
Presence of risky land uses in the catchment (Large scale
0 100 0.03 3.08
7 construcƟons/ Industrials/DeforestraƟon)
8 DistribuƟon of aquaƟc plants in tanks (% Coverage) 95 0 0.02 0.00
30 30 0.21 6.37
1 Existance and Performance of farmer organizaƟons
10 40 0.16 6.44
2 Healthy agricultural pracƟces - organic ferƟlizer usage
20 20 0.12 2.39
3 Percentage of farmers who pracƟce water conservaƟon strategies
4 Percentage of young farmers involved in agricuture 18 40 0.09 3.73
5 MulƟple water uses 64
5.1. IrrigaƟon 1
5.2. Bathing and washing 1
5.3. Drinking 0
5.4. Sand mining 0
5.5. Fishing 1
5.6. CollecƟng firewood from forested areas 1
80 0.04 3.53
Socio-Economic 5.7. HarvesƟng sedges 1
5.8. CollecƟng of bee honey 1 47
Sustainability
5.9. Maintaining the ground water table 1
5.10. HarvesƟg lotus, water lily and other flowers 1
5.11.HarvesƟng Ɵmber 0
5.12. Livestock 1
Score 9
5 PopulaƟon density (PopulaƟon per km 2) 68.43 100 0.10 9.57
6 People's awareness of cascade system funcƟons 63 63 0.07 4.41
7 Women parcipaƟon in cascade acƟviƟes 3 100 0.06
9 Severity of Human Elephant conflict 20 75 0.04 3.36
10 Percentage of Samurdhi beneficiaries 7 75 0.04 2.94
11 Average paddy yield in maha season(kg per ha) 5115 100 0.02 1.59
Cropping Intensity (Total annual irrigated crop area/Total
1 50 0.03 1.56
12 command area)
Number of cascade related projects which has implemented
1 50 0.02 0.78
13 during last 10 years
Fig. 3 Sustainability Index for Village Tank Cascade systems—the Microsoft Excel interface
tanks/cascade systems for different purposes such as tank rehabilitation, project

implementations to obtain optimum benefits from these activities. The index further
interprets the rehabilitation and development needs of the tank/cascade system.
According to the developed sustainability index, it is evident that the contribution
of some of the parameters to the sustainability of the cascade system is signifi-
cantly minimal according to the assessment. In the purpose of rapid assessment of
sustainability or in a situation where there is lack of data, it would be practically

difficult to take all 21 parameters for the assessment. Hence, the parameters which
have higher weights can be used in a rapid assessment. There are four physical-
environmental parameters and four socioeconomic parameters which have more than
10% contribution to the developed sustainability index. These parameters are listed
below:
Physical-Environmental Parameters:
• Ratio of catchment area: Water surface area of tank
• Ratio of live storage of tank: Command area
• Agro-well density in the catchment
• Percentage of forest cover in the catchment
Socio-Economic Parameters:
• Existence and performance of the farmer organization
• Healthy agricultural practices—Organic fertilizer usage
• Percentage of farmers who practice water conservation strategies
• Population density.
These parameters can again be weighted to suit different applications either for
individual tanks or tank cascade systems. Accordingly, the following simplified
equation presents the sustainability of VTCs.
Sustainabilit y o f V T Cs
⎛⎛ ⎞ ⎞
0.44 × Ratio o f catchment ar ea :
⎜⎜ W ater sur f ace ar ea o f tank ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ 0.22 × Ratio o f live storage o f tank : ⎟ ⎟
=⎜⎜ + ⎟×50%⎟
⎜⎜ Command ar ea ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
⎝⎝ +(0.14 × Agr o − well densit y in the catchment) ⎠ ⎠
+(0.20 × Per centage f or est cover in the catchment)
⎛⎛ ⎞ ⎞
0.36 × E xistence and per f or mance o f
⎜⎜ ⎟ ⎟
⎜⎜ f ar mer organi zation ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ ⎟ ⎟
⎜⎜ 0.27 × H ealthy agricultural practices ⎟ ⎟
⎜⎜ + ⎟ ⎟
⎜⎜ −organic f er tili zer usage ⎟ ⎟
+⎜⎜ ⎟×50%⎟ (3)
⎜⎜ ⎟
⎟ ⎟
⎜⎜ ⎟
⎜⎜ 0.20 × Per centage o f f ar mer s who practice ⎟ ⎟
⎜⎜ + ⎟ ⎟
⎜⎜ water conser vation strategies ⎟ ⎟
⎝⎝ ⎠ ⎠
+(0.17 × Population densit y)
4 Conclusions
The developed sustainability index can be used in assessing the sustainability of

any village tank cascade system as well as individual tanks. The sustainability index
presents the overall sustainability as well as physical-environmental sustainability
and socioeconomic sustainability separately. It helps in identifying the rehabilitation
and development needs and comparisons of tanks and cascade systems. Out of 21
parameters considered in the sustainability index, it was identified that four physical-
environmental parameters and four socio-economic parameters contribute more to
the sustainability, and those eight parameters can be used in rapid assessments and
in situations with poor data availability.
Acknowledgements This work was carried out with the aid of a grant from the Interna-
tional Development Research Center (IDRC), Ottawa, Canada. Their financial support is greatly
appreciated.
References
Anuradha, J.M.P.N., Miho Fujimura, Tsukasa Inaoka, and Norio Sakai. 2019. The Role of Agricul-
tural Land Use Pattern Dynamics on Elephant Habitat Depletion and Human-Elephant Conflict
in Sri Lanka. Sustainability 11 (10): 2818.
Bandara, J.M.R.S., D.M.A.N. Senevirathna, D.M.R.S.B. Dasanayake, V. Herath, J.M.R.P. Bandara,
T. Abeysekara, and Kolin Harinda Rajapaksha. 2008. Chronic Renal Failure Among Farm Fami-
lies in Cascade Irrigation Systems in Sri Lanka Associated with Elevated Dietary Cadmium
Levels in Rice and Freshwater Fish (Tilapia). Environmental Geochemistry and Health 30 (5):
465–478.
Bonnesoeur, Vivien, Bruno Locatelli, Manuel R. Guariguata, Boris F. Ochoa-Tocachi, Veerle
Vanacker, Zhun Mao, Alexia Stokes, and Sarah-Lan. Mathez-Stiefel. 2019. Impacts of Forests
and Forestation on Hydrological Services in the Andes: A Systematic Review. Forest Ecology
and Management 433: 569–584.
Campos-Arceiz, Ahimsa, S. Takatsuki, S. Ekanayaka and T. Hasegawa. 2009. The Human-Elephant
Conflict in Southeastern Sri Lanka: Type of Damage, Seasonal Patterns, and Sexual Differences
in the Raiding Behavior of Elephants. Geography 5–14.
Dharmasena, P.B. 2009. Sustainability of Small Tank Irrigation Systems in Sri Lanka at the 21st
Century. Traversing No Man’s Land: Interdisciplinary Essays in Honour of Sadurawan Madhuma
Bandhura 233–252.
Dharmasena, P.B. 2010. Essential Components of Traditional Village Tank Systems. In Proceedings
of the National Conference on Cascade Irrigation Systems for Rural Sustainability, Central
Environmental Authority, Sri Lanka, vol. 9.
Dharmasena, P.B. 2020. Cascaded Tank-Village System: Present Status and Prospects. In Agricul-
tural Research for Sustainable Food Systems in Sri Lanka, 63–75. Singapore: Springer.
Dharmasena, P.B., and I.M. Goodwill. 1999. Use of Groundwater in Minor Tank Irrigation Schemes
of Sri Lanka. In ICID, 17th Congress on Irrigation and Drainage, 175–194. Spain: Granada.
Ekhuemelo, David O., J.I. Amonum, and I.A. Usman. 2016. Importance of Forest and Trees in
Sustaining Water Supply and Rainfall. Nigeria Journal of Education, Health and Technology
Research 8: 273–280.
Geekiyanage, Nalaka, and D.K.N.G. Pushpakumara. 2013. Ecology of Ancient Tank Cascade
Systems in Island Sri Lanka. Journal of Marine and Island Cultures 2 (2): 93–101.
http://www.statistics.gov.lk/agriculture/Paddy%20Statistics/PaddyStatsPages/AverageYield.html.
Retrieved on 16.03.2020.
https://doa.gov.lk/rrdi/index.php?option=com_sppagebuilderandview=pageandid=42andlang=en.
Retrieved on 12.04.2020.
https://www.centreforpublicimpact.org/case-study/samurdhi-programme-sri-lanka. Retrieved on
17.08.2021.
https://www.gwp.org/contentassets/05190d0c938f47d1b254d6606ec6bb04/dublin-rio-principles.
pdf. Retrieved on 17.03.2020.
Ishizaka, Alessio, and Ashraf Labib. 2011. Review of the Main Developments in the Analytic
Hierarchy Process. Expert Systems with Applications 38 (11): 14336–14345.
Jayarathne, K.D.B.L., N.D.K. Dayawansa, and R.P. De Silva. 2010. GIS Based Analysis of Biophys-
ical and Socio-Economic Factors for Land Degradation in Kandaketiya DS Division. Tropical
Agricultural Research 21 (4): 361–367.
Loucks, Daniel, P., and Eelco Van Beek. 2017. Water Resource Systems Planning And Management:
An Introduction to Methods, Models, and Applications. Springer, 1–49.
Madduma Bandara, C. 1985. Catchment Ecosystems and Village Tank Cascades in the Dry Zone
of Sri Lanka a Time-Tested System of Land and Water Resource Management. In Strategies for
River Basin Management, 99–113. Dordrecht: Springer.
Marambe, Buddhi, Gamini Pushpakumara, and Pradeepa Silva. 2012. Biodiversity and Agrobio-
diversity in Sri Lanka: Village Tank Systems. In The Biodiversity Observation Network in the
Asia-Pacific Region, 403–430. Tokyo: Springer.
Oka, Naoko, Takeru Higashimaki, D.D.P. Witharana, and Mekonnen B. Wakeyo. 2015. Constrains
and Consensus on Water Use and Land Allocation in Minor Scheme Tanks in the Dry Zone of Sri
Lanka. Scientific Papers Series Management, Economic Engineering in Agriculture and Rural
Development 15 (4): 185–190.
Panabokke, Christopher R., M. U. A. Tennakoon, and R. de S. Ariyabandu. 2001. Small Tank
Systems in Sri Lanka: Issues and Considerations. In Proceedings of the Workshop on Food Security
and Small Tank Systems in Sri Lanka. National Science Foundation, Colombo, Sri Lanka, 1–6.
Pelpitiya, I.P.S.K., N.D.K. Dayawansa, and E.R.N. Gunawardena. 2016. Impacts of Land Cover
Dynamics and Shallow Groundwater Abstraction on Sustainability of Hakwatuna Oya Irrigation
System. Tropical Agricultural Research 27 (1): 13–26.
Perera, M.P. 2016. The Hydro-Ecological Impact Assessment Scoring: A New Participatory Method
to Grasp Hydro-Ecological Impact of Agro-well Development of Sri Lanka. International Journal
of Advanced Research and Review 1 (7): 01–10.
Sakthivadivel, Ramaswamy, Christopher R. Panabokke, N. Fernando, and C. M. Wijayaratna.
1994. Guidance Package for Water Development Component of Small Tank Cascade Systems.
Colombo, Sri Lanka: International Irrigation Management Institute (IIMI). Sri Lanka Field
Operations (SLFO), for International Fund for Agricultural Development (IFAD).166.
Sakthivadivel, Ramaswamy, Nihal Fernando, Christopher R. Panabokke, and C. M. Wijayaratna.
1996. Nature of Small Tank Cascade Systems and a Framework for Rehabilitation of Tanks Within
Them. Colombo, Sri Lanka: International Irrigation Management Institute (IIMI).54.
Samarasinghe, Y.M.P., and N.D.K. Dayawansa. 2011. Mapping Degradation Risk of Kolonnawa
Marsh: A Study with Remote Sensing and GIS. Undergraduate thesis, Faculty of Agriculture,
University of Peradeniya, Sri Lanka, 50.
Sangrama. 2018. Participatory Rural Appraisal Report: Participatory Climate Risk Vulnerability
Capacity Assessment (pcr-vca) Palugaswewa Cascade, 43.
Santiapillai, Charles, S. Wijeyamohan, Ganga Bandara, Rukmali Athurupana, Naveen Dissanayake,
and Bruce Read. 2010. An Assessment of the Human-Elephant Conflict in Sri Lanka. Ceylon
Journal of Science (biological Sciences) 39 (1): 21–33.
Somarathna, H.M., and H. Kono. 2005. Indigenous Institutions for Irrigation Water Management
and Sustainable Agriculture: A Case Study from Sri Lanka. Journal of Agricultural Development
Studies (Japan), 69–76.
Somaratne, W.G. 2002. Sri Lankan Agriculture for the Next Decade: Challenges and Opportuni-
ties. Hector Kobbekaduwa Agrarian Research and Training Institute, Sri Lanka, 123.
Srdjevic, Bojan, and Zorica Jandric. 2010. Analytic Hierarchy Process in Selecting the Best
Irrigation Method. Agricultural Systems 103 (6): 350–358.
Wijesundara, S. 2010. Invasive Alien Plants in Sri Lanka. Invasive Alien Species—Strengthening
Capacity to Control Introduction and Spread in Sri Lanka. Biodiversity Secretariat of the Ministry
of the Environment and United Nations Development Programme, Sri Lanka, 27–38.
Zondag, Marie-José, Sacha Koppert, Carolien De Lauwere, Peter Sloot, and Andres Pauer. 2015.
Needs of Young Farmers: Report I of the Pilot Project: Exchange Programmes for Young
Farmers. European Commission, Directorate-General for Agriculture and Rural Development,
79.
An Agent Based Model of Mangrove
Social-Ecological System for Livelihood
Security Assessment
Shamima Airin Sweety, M. Shah Alam Khan, Anisul Haque,

and Mashfiqus Salehin
Abstract Ecosystem services and livelihood security are strongly interrelated in

any social-ecological system (SES), and their interdependence should be assessed
using an integrated framework. Overexploitation of ecosystem resources may under-
mine the ecosystem services while restrictions on resource extraction may hamper
the dependent livelihoods. We examine these interrelationships for the mangrove
SES in Bangladesh using an Agent Based Model (ABM), a useful tool for inte-
grated assessment of human behavior and natural processes. The model simulates
the dynamics of ecosystem pressure and livelihood security by linking the supply
and expenditure of ecosystem resources, exploited and/or regulated by five primary
agents, viz. Bawalis, Farmers, and Fishers (livelihood agents), and Forest office and
UP office (institutional agents). We find that the livelihood activities can be consider-
ably influenced, both favorably and adversely, by the ecosystem characteristics and
the institutional efficiencies. Sustainability of the livelihood activities is indicated by
a change in the number of livelihood groups being able to continue their activities
in the long term. This will require sound institutional policy for conservation of the
ecosystem, preventing overexploitation, and supporting production of provisioning
ecosystem services. Supporting optimum levels of livelihood activities and ensuring
sustainability of ecosystem resources will help avert their tipping points.
Keywords Social-ecological system · Livelihood sustainability · Agent based

model · Ecosystem service · Mangrove
S. A. Sweety (B) · M. S. A. Khan · A. Haque · M. Salehin

320 S. A. Sweety et al.
1 Introduction
1.1 Mangrove Social-Ecological System and Ecosystem

Services
Mangroves provide essential ecosystem services to the coastal population and natural
environment. These services include provisioning (e.g., food and water), regulating
(e.g., climate and disease control), supporting (e.g., nutrient cycling and pollination),
and cultural (e.g., recreational and spiritual) dimensions that define the services.
While ecosystem services ensure material and energy flows, leading to economic
flow, human actions and interventions often stress the ecosystem and affect its
services to varying extents. The cultural dimensions influence the way social groups
act and perform. A systems view of the coupled human-natural processes is thus
useful in the analysis of sustainability of the ecosystem services and resilience of the
ecosystem, society, and institutions. This concept of a dynamically coupled human
and natural system (CHANS) can therefore be applied more specifically to an inte-
grated social-ecological system (SES) of the mangrove ecosystem and its dependent
population.
The SES framework emerged from the concepts of common pool resources and
integrated social-ecological processes (Ostrom 1990; Ostrom et al. 2007). A SES is
an integrated and adaptive system representation of the social and ecological compo-
nents and processes and broadly represents the human and natural domains of the
system (see Fig. 1). The complex dynamics of the integrated system, particularly in
different dimensions and scales, however, are yet to be fully understood and oper-
ationalized through analytical methods and tools. Virapongse et al. (2016) present
a framework to translate the SES theory into practice for improved environmental
management by defining the key components and processes in the system. Interac-
tion of processes in the human and natural domains of the SES gives rise to a set of
management decisions and adaptation choices, while the system and its sub-domains
Fig. 1 Components and

processes in a
social-ecological system
[Adapted from Virapongse
et al. (2016)]
An Agent Based Model … 321
are influenced by external factors such as political, economic, and biogeochemical

conditions. Other approaches to operationalization of the SES theory and represen-
tation of the nonlinear system properties and dynamics are found elsewhere in the
literature (Cole et al. 2019; Forrester et al. 2014; Schlüter et al. 2012, Schlüter et al.
2014).
1.2 Livelihood Security and Dependency
The Sundarbans mangroves are rich sources of livelihood resources such as timber,
mangrove palm, fish, shrimp, crab, and honey. Rice is also grown in the fringe areas of
the mangroves. Household income and food security of the communities living in and
around the mangroves depend mostly on these resources. Although the households
are clustered and known by their main livelihood activities, household members
are often engaged in multiple livelihoods to maximize the household income. Also,
different livelihood groups in a community are connected and interdependent through
various social and economic activities.
Local-level institutions, such as the Forest Office, regulate resource extraction to
prevent overexploitation of resources and preserve ecosystem health and undertake
other measures for ecosystem conservation and growth. Other institutions, such as
the Union Parishad, the lowest tier local government institution, provide support
to the livelihood groups through subsidies, training, and awareness campaign. Thus,
sustainability of the mangrove-dependent livelihoods depends not only on the natural
resources, but also on the institutional policies and broader economic, political, and
environmental conditions.
1.3 SES Tipping Points
Sustainability of both the ecosystem and the dependent livelihoods is ensured when
the resource extraction and ecosystem growth are balanced. This equilibrium can
be achieved through regulatory policy and participatory decision making. However,
environmental hazards or overexploitation often leads to negative feedback loops
in the system forcing it to reach a tipping point. For example, overexploitation of
fish increases household income, but rapidly depletes the fish availability, eventu-
ally leading to an overall decline of the SES sustainability. The tipping point is the
‘point of no return’ beyond which the SES can no longer be brought back to the
previous equilibrium even by lowering the fish catch. Examples of SES feedback
loop, ecosystem service trade-off, food security, and livelihood sustainability are
given in Miyasaka et al. (2017), Acosta et al. (2018), Dobbie et al. (2018), and Yan
et al. (2019).
1.4 Agent Based Modeling for SES Analysis
Agent based models (ABM) can help develop a simplified understanding of the
complex and dynamic interaction among the ecosystem components, human actors,
and institutions. This understanding is essential for sustainable environmental
management and livelihood security. ABMs idealize the SES as a composition of
autonomous, decision-making agents that are able to follow their goals, based on a
set of rules. These agents may range from plant species to fish and animals, to house-
holds, livelihood groups, and institutions (Schulze et al. 2017; Rounsevell et al. 2012;
Macal and North 2009). ABMs can also analyze the complexities such as hetero-
geneity, nonlinearity, and feedback in a CHANS or integrated SES, support geospatial
simulation or participatory modeling of the SES, explore the system behavior across
scales, and clarify the system dynamics in varying conditions (An 2012; An et al.
2014; Castle and Crooks 2006; Voinov et al. 2018; Lippe et al. 2019; Martin and
Schlüter 2015; Hossain et al. 2017).
In this paper, we present an ABM application to analyze the mangrove SES-based
livelihood processes, simulate the dynamics of livelihood activities and ecosystem
responses, determine the efficiency of institutions in SES management, and explore
how livelihood decisions and tipping points vary in different scenarios. In partic-
ular, we assess the ecosystem pressure in response to livelihood activities and try to
understand how restrictions on resource extraction impact livelihoods.
2 Study Area and Methods
2.1 Study Area Description
2.1.1 SES Components and Linkages
Gabura union under Shyamnagar upazila of Satkhira district and its surrounding
areas of the Sundarbans is selected as the study area (Fig. 2). Since the life and
livelihoods of people in Gabura are strongly dependent on the mangrove resources,
the study area is an excellent example of a SES. The mangrove forest (Sundarbans),
of its total of 10,000 km2 , covers an area over 6,017 km2 in Bangladesh. This world
heritage site provides natural protection against cyclones and storm surges. The
fertile soils of this area have been used intensively in agriculture for centuries, and
the fringes of this ecoregion have been converted for intensive agriculture thereby
decreasing the forest area alarmingly. This transition region between freshwater from
the Ganges and saline water of the Bay of Bengal provides a brackish environment
rich in biodiversity.
Golpata or mangrove palm (Nypa fruticans) is a very common palm species of
the Sundarbans which grows in abundance here naturally, mainly in the less saline
zone but can also grow in the moderate saline zone. Golpata is of high demand as
Fig. 2 Location and physical features of the mangrove social-ecological system under study
an economically important product for its widespread domestic and commercial use.
However, the golpata stock in the Sundarbans has been severely decreasing because
of both licensed and unauthorized extraction in excessive amounts, and due to the
lack of proper forest management.
Fishes of the Sundarbans represent 322 species belonging to 217 genera, 96 fami-
lies, and 22 orders (Habib et al. 2020), with at least 150 species of commercially
important fish (Wikipedia 2021). The forest also supports the habitat for econom-
ically significant mud crabs or mangrove crabs (Banglapedia 2021). Fishing and
collecting fish and crab seedlings from the mangrove and nearby offshore waters
for fish farming are very common livelihoods of the people living in and around the
Sundarbans. Many people are also engaged in fish farming and household fishing.
Overfishing and excessive seedling collection to meet the increasing domestic and
commercial demands has become a threat for the sustainability of the ecosystem
services of the Sundarbans (Hoq 2007).
2.1.2 Physical and Hydro-Climatic and Settings
Gabura union is situated on the bank of the Kholpetua River, which separates this
union from the Sundarbans (Fig. 2). The Kopotakkho River flows on the east side of
this union, bordering with Dakshin Bedkasi union of Koyra upazila of Khulna district.
Gabura is also adjacent to Padmapukur union on the north and is 27 km and 82 km
away from the upazila headquarter and Satkhira district headquarter, respectively.
Gabura is one of the few areas near the Sundarbans that have been frequently
affected by cyclones and storm surges. Climate change is believed to have increased
the frequency of these high-energy cyclones. Fresh water availability has been
severely limited in Gabura, which becomes acute after every cyclone. The brackish
water is useful for the fish farms, but adversely affects crop production and fresh-
water habitats. Criss-crossed with a network of rivers and canals, the area suffers
from sediment deposition in the water bodies.
2.1.3 Primary and Alternative Livelihoods
The major livelihoods in this area include crab fattening, shrimp farming, agriculture,
fishing, golpata and honey collection, day labor, van/auto bike pulling, and small busi-
nesses. About 42% of the people in the study area are involved in aquaculture, espe-
cially crab fattening for their livelihoods, while approximately 19% are dependent
on forest resources, 16% are involved in shrimp farming, and 23% engage in agri-
cultural activities, mainly paddy cultivation (http://www.gaburaup.satkhira.gov.bd).
The primary mangrove-dependent livelihoods are fishing in mangrove water bodies
and household ponds, extracting golpata and paddy cultivation. Golpata extraction is
possible only in winter time in mid-March to mid-May; therefore, the Bawalis, i.e.,
the golpata collectors, do fishing as an alternative livelihood. Mangrove-dependent
fishers of this area go into the mangroves for fishing, with a small group of them also
doing fishing in household ponds. These fishers do not take other livelihood options
although they are required to stop fishing in the mangroves due to fishing ban for one
month of the year usually in July. Other fishers do fishing in household ponds only.
2.2 Agent Based Model for Mangrove SES
The important steps for developing an ABM for the Mangrove SES include selecting
important agents, developing a conceptual model delineating how the agents interact
with each other and their environment, and defining the rules; mathematically how
the agents would behave, interact, and take decisions autonomously. The conceptual
framework and the rules for different agents were based on review of secondary
literature on Sundarban-based livelihood systems (e.g., Mallick et. al. 2021; Kabir
et. al. 2019; Mozumder et. al. 2018; Getzner and Islam 2013; Sarker 2011 and
https://www.ccec-bd.org › files › reports), direct mobile communication with few
representatives of the primary agents (e.g., fishermen, Bawalis), and expert judgment.
Some baseline data was collected from the Bangladesh Bureau of Statistics (BBS)
database (http://www.bbs.gov.bd/).
The model can be considered exploratory at this stage. The conceptual model and
the rules for agent behavior will require detailed verification and validation in the
field with the primary agents (which was limited during this study because of the
prevailing COVID situation) before the ABM for the mangrove SES can be used as
a fully robust model. Nevertheless, the approximations made in the formulation are
considered reasonable, if not fully accurate, and hence the model should be able to
provide useful early insights.
Table 1 Agents and model

Class of Agents Name of Agents Model Environment
environment
Livelihood Agent Bawalis Mangrove
Fishers Socio-Ecological
System
Farmers
Institutional Agent Union Parishad
Office (UPO)
Forest Office
(FO)
2.2.1 Agent Selection for ABM
Agents are individual actors or actor groups of a SES who can interact with each other
and their environment by autonomous decision-making based on specific rules. This
study simulated the interrelations of the main mangrove-dependent livelihood groups
(both individual actors and actor groups) with each other and with the environment
of the SES. The livelihood activities are also linked with the regulations by the Union
Parishad Office (UPO) and Forest Office (FO). For this reason, three main livelihood
groups of this study area: Bawalis, Fishers, and Farmers, and these two institutions
have been considered as the agents in the ABM model. The agent classification and
model environment are given in Table 1.
2.2.2 Conceptual Model
Following the agent descriptions and attributes discussed earlier, Fig. 3 shows a
conceptual model of interactions among the agents of the mangrove SES. Golpata is
Fig. 3 Conceptual model of the mangrove social-ecological system showing the agents and their
livelihood interactions
a primary source of livelihoods of the Bawalis. The Bawalis mainly extract golpata
from the Sundarbans with a permit from the FO to extract a maximum fixed amount.
The extraction amount of the Bawalis depends on their extraction capacity and the
permitted extraction amount. The FO gives permission till the golpata stock of the
patches in the forest reaches a threshold maximum amount after which the permission
ceases. Also, the Bawalis need to attain a threshold capacity to go to the forest
for extracting golpata even if they get a permit. Otherwise, they mainly do fishing
or farming to support household activities and increase their extraction capacity
gradually. Golpata grows and dies in a natural way at a net growth rate, while the
golpata stock decreases due to natural hazards at an average rate. For each extraction
trip, the Bawalis need to spend a portion of resources, which is subtracted from their
actual extraction amount. A fraction of the remaining extracted amount contributes
to increasing the extraction capacity for the next trip. Rest of the extracted amount
is spent for household living and other purposes. The FO stops giving permits if
the golpata stock reaches a minimum threshold amount when the FO also starts
conservation activities for golpata to grow rapidly.
Two types of fishers are considered in this study. The first type of fishers mainly
goes to the mangroves to catch fish using boats after getting a permit from the FO.
The FO generally stops giving fishing permits for one month per year to support fish
breeding, during which time these fishers remain jobless. For catching fish, they need
to spend a portion of resources which is subtracted from their catching capacity in
terms of the catch amount. A fraction of the catch amount contributes to increasing
the catching capacity for the next year, which is proportional to the number of fish
caught in the current year. Rest of this amount is spent for household living and
other purposes. They can take permits from the FO many times in a year, with each
permit valid for one week. Although the catch amount during a trip is uncertain,
these fishers can generally catch a good amount for 7 months, a satisfactory amount
for 2 months and small amounts for the rest of the year. Their expenditure, mainly on
living, slightly increases if they can achieve a certain threshold of catching capacity.
If their catching capacity goes down below a minimum threshold, they need to take
loans for continuing fishing. They must pay back the loan with interests at a higher
rate. Some of these fishers also do fishing in their household ponds.
The second type of fishers only grows fish in their own household ponds. They
need to bear the production cost which is subtracted from their catching capacity. A
fraction of the production increases the catching capacity for the next year and the
remaining amount is used for household living and other purposes. The procedures
for taking loan and repayment are similar to the first type of fishers. Fishes in the
mangroves and household ponds increase due to breeding and decrease due to death
and catching. Besides, if the fishes increase at a certain maximum threshold rate, a
certain population of the fish dies naturally.
Farmers grow crops (mainly paddy) according to their crop production capacity
and crop productivity of the agricultural land. They need to bear fertilizer cost to
produce paddy which is subtracted from their crop production capacity. Crop produc-
tivity of the land decreases due to natural hazards, mainly soil salinization from
storm surges. The actual production amount of paddy depends on the remaining crop
production capacity and crop productivity. A fraction of this actual amount increases
the crop production capacity of the farmers for the next year. The remaining amount
is used for living and other expenses at the household level. Farmers need to achieve
a minimum crop production capacity to produce paddy; otherwise, they have to take
loans. The loan is repaid later at a higher rate with interest.
The UPO performs regulatory functions in the study area to support financial
security of the livelihood groups. It also coordinates with the FO for forest conser-
vation. The efficiency of providing financial security depends on the capacity of
livelihood activities of the livelihood groups. For example, if the number of active
Bawalis decreases to a certain minimum, it indicates the lack of efficiency of the
UPO in supporting their financial security. Similarly, if the number of fishers and
farmers who have loans to pay increases to a certain maximum, it indicates a lower
efficiency of the UPO in supporting their financial security.
The FO regulates extraction in mangroves for mangrove protection and conser-
vation. When the natural hazard loss exceeds a certain amount, and the conserva-
tion growth rate remains low at the same time, it indicates low forest conservation
efficiency. Since the livelihood groups depend on ecosystem resources directly for
their livelihood activities, they produce pressure on the ecosystem. Such pressures
produced by each livelihood group are calculated from the Net Primary Productivity
(NPP) (from the net carbon gain) supply of the ecosystem and the NPP consumed
by the livelihood group.
2.2.3 Mathematical Descriptions
Mathematical formulation of the model follows an agent-wise rule-based approach

where the agents follow their own procedures according to the rules and behave
autonomously while they interact with each other and with their environment. The
pressures produced by the livelihood groups are calculated from the NPP supply on
the ecosystem and the NPP consumed by each livelihood group where the NPP is
estimated from the net carbon gain. The multipliers with variables and subtractors
from variables in the equations are taken hypothetically considering logic where
multipliers are numbers and subtractors have the same unit of associated variables.
Rules Related to Bawalis
The rules for the Bawalis are defined in terms of their extraction capacity, actual
extraction amount, new actual extraction amount, golpata permit, golpata stock,
their cost of movement, natural growth rate of golpata, conservation growth rate of
golpata, and natural hazard loss of golpata. All these components are expressed in
Metric tons.
1. If permit for golpata extraction remains active, then the Bawalis need to have a
minimum capacity to extract golpata and they can extract the amount of golpata
according to the amount permitted by the FO and their capacity for extraction.
If their capacity is higher than the permitted amount, they need to extract the
amount equal to the permitted amount, otherwise they can extract equal to their
capacity which is their actual extraction amount.
2. They need to spend for their movement to the mangroves and it reduces some
units of their actual extraction amount as follows:
new actual extraction amount = actual extraction amount

− movement cost of bawali (1)
3. A fraction of this increases the extraction capacity of Bawali for the next year.
So, their increased capacity becomes:
extraction capacity = extraction capacity

+ (0.01 × new actual extraction amount) (2)
4. If the permit remains off and they have the minimum necessary capacity for
extraction, then they lose some units of their extraction capacity as follows:
extraction capacity = extraction capacity − 10 (3)
5. If they do not have the minimum capacity, then they increase their capacity by
some units for extraction in the next year by doing fishing or farming as follows:
extraction capacity = extraction capacity + 2 (4)
6. If golpata stock of patches has a certain maximum unit of golpata, then the
golpata stock changes by the following equation:
golpata stock = golpata stock + golpata natural growth rate − natural hazard loss
(5)
7. If the golpata stock of patches reaches at a certain minimum amount, FO starts

conservation. In this case,
golpata stock = golpata stock + natural growth rate

+ conservation growth rate − natural hazard loss (6)
Rules Related to Fishers
Rules for the fishers are defined in terms of their apparent and actual catching capacity,
amount of loans they take, cost of ice for fish preservation, their movement cost, and
fish production cost. All these components are expressed in Metric Tons.
1. Fishers who catch fishes in mangrove water area need to spend movement cost
for fishing in mangroves and for carrying ice to protect fishes. It decreases
some units of their capacity. So, their actual catching capacity becomes:
actual catching capacity = catching capacity − movement cost − ice cost

(7)
2. If they can catch a good amount of fishes for at least seven months per year, it
increases their catching capacity by some units as follows:
catching capacity = catching capacity + (0.05 × actual catching capacity)

(8)
Otherwise,
catching capacity = catching capacity − 0.075 (9)
3. If they can catch a moderate amount of fishes for at least two months per year,
it increases their catching capacity by some units as follows:
catching capacity = catching capacity + 0.075 (10)
Otherwise,
4. If they can catch a small amount of fishes for a maximum of two months per
year, it increases their catching capacity by some units as follows:
Otherwise,
5. Fishing in mangroves remains stopped for one month, and the fishers become
jobless at this time. In this case,
6. If their catching capacity becomes greater than 2.5 Metric ton, they increase
additional expenses and therefore:

7. If their catching capacity becomes less than 2 Metric ton, they need to take loan
to increase some units of their catching capacity by the following equation,
8. After achieving a catching capacity greater than 2.5 Metric ton after taking
loan, they pay back their loan at a slightly increased amount, which is given
by the following equation:
9. Some of these fishers also do household fishing and thus increase some units
of their catching capacity for fishing in mangroves by,
10. Fishers who grow fishes in household ponds need to spend a production cost
and therefore their actual catching capacity becomes:
catching capacity = catching capacity − fish production cost (19)
11. If the production is good for most of the year, then their catching capacity
increases as:

(20)
12. If their fish production for the year is moderate, then:

(21)
13. If they produce a small amount of fishes in the year, they cannot increase their
catching capacity.
14. If they produce a very low amount of fish which is not noticeable, their catching
capacity decreases as follows:
15. If their catching capacity becomes greater than 2.7 Metric ton, they increase
their other expenditure, and it reduces some units of their catching capacity as:

16. They need to take loan, if their catching capacity becomes less than 2 Metric
ton, to increase their catching capacity as:
17. They pay back the loan after gaining a catching capacity greater than 2.5 Metric
ton, which is represented by,
Rules Related to Farmers
Rules for the farmers are defined in terms of crop production capacity, actual produc-
tion amount, loan, crop productivity, actual crop productivity, fertilizer cost, and
natural hazard loss of crops. All these components are expressed in Metric Tons.
1. Farmers’ actual crop production depends on the actual crop productivity of their
agricultural land and their own capacity. They need to also pay for fertilizer.
The crop productivity of agricultural land decreases due to natural hazards. If
their capacity is higher than half of the actual crop productivity of the land, their
actual production amount then becomes half of the actual crop productivity of
the land (since it is observed that two farmers produce in each land); otherwise,
they produce the amount to their capacity.
2. They need a certain minimum capacity to produce crops.
3. If the actual crop productivity is greater than 15 Metric ton, then their capacity
increases as:
crop production capacity = crop production capacity

actual production amount
+ 0.1 × (26)
−fertilizer cost
4. If the actual crop productivity is between 10 Metric ton and 15 Metric ton, the
crop production capacity increases as:
crop production capacity = crop production capacity

actual production amount
+ 0.05 × (27)
−fertilizer cost
5. If the actual crop productivity is less than 15 Metric ton, then crop production
capacity becomes:
crop production capacity = actual production amount − fertilizer cost (28)

6. They take loan if their crop production capacity is less than or equal to 2.5
Metric ton which is given by,
crop production capacity = crop production capacity + 1 (29)
7. They pay back the loan when they regain their crop production capacity to
greater than or equal to 4 Metric ton as below:
crop production capacity = crop production capacity − 1.25 (30)
Rules Related to UPO
1. If the mean extraction capacity of the Bawalis, mean catching capacity of both
types of Fishers, and mean crop production capacity of Farmers become less
than 125, 2.3, and 4.5 Metric ton, respectively, and the number of active Bawalis
becomes less than half of the total number of the Bawalis, then financial security
efficiency (a dimensionless quantity or score) of the UPO will be 0.5; otherwise,
it will be 1 for each of these 5 cases.
2. If the number of Fishers of both types and Farmers who have loans to pay is
greater than their total number for each of these three types of livelihoods, then
financial security efficiency of UPO will be 0.5; otherwise, it will be 1 for each
of these 3 cases.
3. The efficiency of UPO is calculated by,
efficiency of UP Office = regulatory efficiency

+ support of financial security efficiency
+ financial safety efficiency (31)
Rules Related to FO
1. If for natural hazard, the loss of golpata becomes greater than 1 Metric ton, and
in this case conservation growth rate of golpata remains less than 4 Metric ton,
the forest conservation efficiency (dimensionless quantity or score) of the FO
will be 2; otherwise, it will be 4.
2. The efficiency of FO is given by,
efficiency of Forest Office = extraction control efficiency

+ forest conservation efficiency (32)
Pressure on the Ecosystem
Pressure on the ecosystem is calculated by the following equation for all livelihood
groups (Yan et al. 2019),
NPP consumed
pressure on ecosystem = (33)
NPP supply
Pressures on ecosystem are numbers as they are ratio of the amount of carbon.
Simulations using NetLogo
Based on the conceptual framework for the ABM, mathematical equations were
set for formulating the computational framework. The NetLogo software (https://
ccl.northwestern.edu/netlogo/) was used for simulations of the ABM (the interface
is shown in Fig. 4). NetLogo is a multi-agent programming language and modeling
environment for simulating complex systems evolving over time (Tisue and Wilenski
2004).
NetLogo visualizes the model by a ‘NetLogo world’ (made up of agents). It has
four types of agents:
Turtles: This type of agents can move around in the world.
Fig. 4 NetLogo ‘Interface’ of the model

Patches: This type of agents has fixed coordinates, and turtles can move over
patches.
Links: Links are used to connect two turtles. Links are of two types—directed
and undirected.
Observers: They do not have locations and only give instruction to other agents.
Four livelihood breeds: Bawalis, Fishersmen_M (for mangrove fishermen), Fish-
ermen_H (for household fishermen), and Farmers; and two institutional breeds: UPO
and FO are used in the NetLogo model. Three types of fish breeds for the mangrove
area and three types of fish breeds for household ponds are included to represent
different amounts of catches during a year. These breeds are referred to as ‘turtles’
in NetLogo as defined earlier. The model uses four types of patch variables, which
include golpata permit, golpata stock, crop productivity, and actual crop productivity.
Global variables are used in the model controlled by the sliders (to define different
scenarios, e.g., favorable, moderate, and critical), which include movement cost of
the Bawalis, golpata growth rate (natural and conservation rates), loss of golpata,
fish, and crops due to natural hazards, cost of ice for the fishermen, movement cost
of fishermen, production cost of fish, fertilizer cost for the farmers, extraction control
efficiency of the FO, and the regulatory efficiency of the UPO.
3 Results
3.1 Cases and Scenarios
Three cases of the SES are selected to explore the system performance and agent
behavior within a feasible range of variables. These cases represent the variation in
catching capacity of the two types of Fishers, crop production capacity of Farmers,
and permitted amount of golpata for the Bawalis. The model variables which define
the differences among the three cases are described in Table 2. Case 1 represents
comparatively high golpata permit for the Bawalis, comparatively low catching
capacity for both mangrove and household fishers, and comparatively high crop
production capacity, while Case 2 and Case 3 represent moderate and comparatively
low golpata permits for the Bawalis, moderate and comparatively high catching
capacities for both mangrove and household fishers, and moderate and comparatively
low crop production capacity, respectively.
Additionally, different global variables also influence agent behavior; some influ-
ence positively while others cause negative effects. These variables also influence
how the ecosystem responds positively or negatively in different conditions.
Three scenarios are selected for simulation in each case: (i) a favorable scenario
considering the highest values in the NetLogo interface slider for the global variables
with positive impact and the lowest values of global variables with negative impact;
(ii) a moderate scenario considering moderate values of all these global variables;
(iii) a critical scenario where the highest values of global variables in the slider with
Table 2 Variable values for

Case Number Variable Value (Metric ton)
the three cases
Case 1 golpata_permit 125
catching_capacity_M 2.3
catching_capacity_H 2.3
crop_production_capacity 5
crop_production_capacity 4.75
crop_production_capacity 4.5
Table 3 Global variable values for the three scenarios

Global Variable Scenario Unit
Favorable Moderate Critical
movement_cost_bawali 10 14.75 20 Metric ton
golpata_conservation_growth_rate 150 125 100
golpata_natural_growth_rate 75 62.5 50
natural_hazard_loss 50 75 100
ice_cost 0.25 0.375 0.5
movement_cost_fishermen_M 0.25 0.375 0.5
production_cost_fish 0.25 0.5 0.75
natural_hazard_loss_crops 0.5 0.75 1
fertilizer_cost 0.5 0.75 1
extraction_control_efficiency 4 3 2 -
regulatory_efficiency 2 1.5 1
negative impact and the lowest values of global variables with positive impact are
selected. These values of the global variables are given in Table 3.
3.2 Agent Behavior in Different Conditions
NetLogo simulations of the varying behavior of the agents in different cases and
scenarios are shown in Figs. 5, 6, 7, 8 and 9. In Netlogo, time is arbitrarily represented
as tics, whereas one tic is equivalent to a year in this model. The figures show how
(a) Bawalis change of behavior in favorable scenario
(b) Bawalis change of behavior in moderate scenario
(c) Bawalis change of behavior in crical scenario
Fig. 5 Behavior change of Bawalis. Case 1: comparatively high golpata permit (Metric ton); Case
2: moderate golpata permit (Metric ton); and Case 3: comparatively low golpata permit (Metric ton)
the agents perform in different ways in different conditions, i.e., one set of variables
or condition may be favorable for certain agents while the same condition may be
unfavorable for the others. Individual agent behaviors are shown in these plots.
Figure 5 shows the behavior of Bawalis in different conditions. The initial extrac-
tion capacity of the Bawalis is only 127 metric ton in all scenarios for case 1, case 2,
and case 3 with permit for golpata extraction 125, 120, and 115 metric ton respec-
tively but in favorable scenario, increasing rapidly with time it reaches at almost 150,
142, and 140 Metric ton at the 100th year for case 3, case 2, and case 1 respectively.
Above 365 bawalis can be active in most of the times for case 3 with extraction
permit 115 Metric ton which is the best among all considered cases in this scenario.
The Moderate scenario has also good extraction capacity ranges from almost 116
to 118 Metric ton and number of active Bawalis ranges from 180 to 240 in most of
the times according to different cases. The best performance of this scenario is for
case 3 where extraction capacity is 117 Metric ton and the number of active Bawalis
207 are almost constant in most of the times with lowest golpata permit of the three
cases, which allows better conservation of golpata stock, and as such relatively high
number of Bawalis can remain active for extraction in future. In critical scenario,
almost constant extraction capacity 114 Metric ton and the number of active Bawalis
below 100 in all cases indicate critical situation of Bawalis.
Figure 6(A) indicates that the catching capacity of the mangrove fishers remains
constant with value above 2.45 Metric tons for all cases and scenarios. The number
of this type of fishers who have to pay loan increases gradually with time, for a long
period and then becomes constant. This implies that the mangrove fishers struggle
initially with repaying loan, but eventually overcome this with the attainment of
stable catching capacity. The number is, however, the highest in the critical scenario
of Case 2 (ranges from 122 to 641) and lowest in the favorable scenario of Case 2
(ranges from 123 t0 468) where catching capacity is moderate to comparatively high.
In moderate scenario, the number of this type of fishers to pay loan has values ranging
from 119 to 552, comparatively lower than the critical scenario and comparatively
higher than the favorable scenario. This number in different times of the moderate
scenario is the lowest for case 3 of all the cases and remains constant after the 80th
years. Figure 6(B) indicates that catching capacity of household fishers for all cases
does not change significantly with the scenarios (ranges between 2.55 and 2.6 Metric
ton) but changes with time, and this value remains constant most of the times for
case 3 in favorable and moderate scenarios. The number of this type of fishers having
loan remains very low in all scenarios (ranges from 1 to 8).
Figure 7 represents the behavior change of farmers for case 1, case 2, and case 3
with crop production capacity 5, 4.75, and 4.5 Metric ton respectively in favorable,
moderate, and critical scenario. From this it is noticed that crop production capacity
of farmers remains the highest (above 4.7 Metric ton) in the favorable scenario in Case
3 with lowest number (below 400) of farmers having loan to pay of all conditions.
In the critical scenario, crop production capacity indicates critical values (3.8 to 4.2
Metric ton) in most of the times for all cases. The number of farmers who have to
pay loan is the lowest (below 450) in the favorable scenario and highest (above 500)
in critical scenario for all cases for most of the times. The Moderate scenario of all
cases indicates moderate number (about 500 to 525) of such farmers which does
change significantly with time. In this scenario, Case 3 indicates the highest catching
capacity whose values are above 4.3 Metric ton with lowest number of farmers having
loan to pay whose maximum value is 500 in different times.
Figure 8 shows the efficiency change of FO and UPO for Case 1 with extraction
permit 125, catching capacity 2.3 and crop production capacity 5 metric ton respec-
tively, Case 2 with extraction permit 120, catching capacity 2.4 and crop production
capacity 4.75 metric ton respectively, and Case 3 with extraction permit 115, catching
capacity 2.5 and crop production capacity 4.5 metric ton respectively in favorable,
moderate, and critical scenario. Figure 8(A) shows that the FO works with the best
efficiency 8 in the favorable scenario. The efficiencies are comparatively lower in
the moderate and critical scenarios which are 7 and 6, respectively. This efficiency
(a) fishermen_M change of behavior in favorable scenario
(b) fishermen_M change of behavior in moderate scenario
(c) fishermen_M change of behavior in cri cal scenario

(B) Behavior changes of fishermen_H
(a) fishermen_M change of behavior in favorable scenario
(b) fishermen_M change of behavior in moderate scenario
(c) fishermen_M change of behavior in cri cal scenario
Fig. 6 Behavior change of Fishers. Case 1: comparatively low catching capacity; Case 2: moderate
catching capacity; Case 3: comparatively high catching capacity
(a) farmers change of behavior in favorable scenario
(b) farmers change of behavior in moderate scenario
(c) farmers change of behavior in cri cal scenario
Fig. 7 Behavior change of Farmers. Case 1: comparatively high crop production capacity (Metric
ton); Case 2: moderate crop production capacity (Metric ton); Case 3: comparatively low crop
production capacity (Metric ton)
does not change according to cases. Figure 8(B) shows variations in efficiency of the
UPO with different conditions. The efficiency is the best ranges from (9.5 to 10) in
the favorable scenario. However, the efficiency remains at a reasonably satisfactory
and constant level in the moderate scenario of Case 3 (ranges from 7.5 to 9). This
efficiency is the lowest in the critical scenarios for all cases and has a value of 6.5
most of the times.
(A) Behavior changes of FO (B) Behavior changes of UP Office
(a) Change of behavior of FO in favorable scenario (a) Change of behavior of UP Office in favorable
scenario
(b) Change of behavior of FO in moderate scenario

(b) Change of behavior of UP Office in moderate
scenario
(c) Change of behavior of FO in cri cal scenario
(c) Change of behavior of UP Office in cri cal scenario
Fig. 8 Behavior change of institutional agents: (A) Forest Office; (B) UP Office
3.3 Ecological and Livelihood Tipping Points
Figure 9 shows the pressure on the ecosystem due to livelihood activities, and how
the system reaches at an equilibrium condition (stable condition) or a tipping point
(unstable condition) in different cases and scenarios. Figure 9(A) shows that in the
favorable scenario of Case 3, the pressure on the ecosystem caused by the Bawalis
increases rapidly, and after the 30th year its value remains almost constant with
time. This value has the highest range (0.12 to 0.14), moderate range (0.06 to 0.08)
in favorable scenario and moderate scenario, respectively. In critical scenario, this
value remains very low most of the times but changes significantly with time. 9(B)
shows that the mangrove fishers cause pressure constantly with time and cases. The
pressure is the lowest (below 0.17) in the critical scenario and the highest (above
0.21) in the favorable scenario, the moderate scenario indicates value almost 0.18.
Figure 9(C) shows that the household fishers cause the minimum pressure in the
(A) Bawalis
(a) Changes in pressure of Bawalis in favorable scenario (b) Changes in pressure of Bawalis in moderate scenario (c) Changes in pressure of Bawalis in crical scenario
(a) Mangrove fishers
(a) Changes in pressure of fishermen_M in favorable scenario (b) Changes in pressure of fishermen_M in moderate scenario (c) Changes in pressure of fishermen_M in crical scenario
(b) Household fishers
(a) Changes in pressure of fishermen_H in favorable scenario (b) Changes in pressure of fishermen_H in moderate scenario (c) Changes in pressure of fishermen_H in crical scenario
(c) Farmers
(a) Changes in pressure of farmers in favorable scenario (b) Changes in pressure of farmers in moderate scenario (c) Changes in pressure of farmers in crical scenario
Fig. 9 Pressure on the ecosystem in different conditions caused by: (A) Bawalis; (B) Mangrove
fishers; (C) Household fishers; and (D) Farmers
critical scenario (below 0.35) and the maximum pressure in the favorable scenario
(above 0.4) where this pressure is about 0.35 in moderate scenario. The pressure
remains constant with time and cases. Figure 9(D) shows that the pressure caused by
the farmers remains the highest for Case 3 (almost 0.6) in the favorable scenario for
a long period of time. The pressure remains consistently moderate in Case 3 (ranges
from 0.51 to 0.53). This pressure is about 0.5 for all cases of critical scenario.
4 Discussion
4.1 Sensitivity Analysis
From the NetLogo simulation results, it is clear that moderate scenario of Case 3 is
the best condition for livelihood activities by keeping balance with ecosystem pres-
sure where the favorable scenario gives the best livelihood performance by creating
excessive pressure on the ecosystem, and the critical scenario creates the minimum
pressure on the ecosystem by giving poor livelihood performance. In this regard, we
have tried to find out the conditions when livelihood performance will be better
than moderate scenario, creating lower pressure than favorable scenario consid-
ering case 3. To do this we have taken the same values for all global variables
as considered for moderate scenario except only four most sensitive global vari-
ables and considered 2 other conditions under case 3 which are case 3A and case 3B.
Since golpata_conservation_growth_rate has positive effects on livelihood activities,
increased values than the moderate scenario of case 3 (i.e., case 3_moderate) at the
same rate have taken for case 3A and case 3B respectively. Again, movement_cost
and fertilizer_cost have negative effects and therefore their decreased values than case
3_moderate have taken for case 3A and case 3B. The only global variable in this case
related to fishermen_H is production_cost_fish. So, increased and decreased values
at the same rate from case 3_moderate have taken for case 3A and case 3B respec-
tively. The detailed values with variables name are given for different conditions in
the Table 4.
Figure 10 shows the changes of behavior of agents for different conditions with
time comparing with the moderate scenario of case 3 (case 3_moderate). From
Fig. 10(A), it is shown that the extraction capacity and number of active bewail
Table 4 Global variable values for the three conditions

Global variables case 3A case 3_moderate case 3B Unit
movement_cost_bawali 14.75 14.75 14.75 Metric ton
golpata_conservation_growth_rate 137.5 125 150
golpata_natural_growth_rate 62.5 62.5 62.5
natural_hazard_loss 75 75 75
ice_cost 0.375 0.375 0.375
movement_cost_fishermen_M 0.3125 0.375 0.25
production_cost_fish 0.625 0.5 0.375
natural_hazard_loss_crops 0.75 0.75 0.75
fertilizer_cost 0.625 0.75 0.5
extraction_control_efficiency 3 3 3 -
regulatory_efficiency 1.5 1.5 1.5 -
(a) bawalis change of behavior
(b) fishermen_M change of behavior
(c) fishermen_H change of behavior
(d) farmers change of behavior
(e) Change of behavior of FO and UPO
Fig. 10 Behavior change of agents: (A) Bawalis; (B) Mangrove fishers; (C) Household fishers; and
(D) Farmers
increase rapidly with time for the case 3A and case 3B than case 3_moderate condi-
tion and case 3B give the best performance with extraction_capacity 122 Metric ton
and 225 number of active bawali most of the times. Figure 10(A) represents that the
catching capacity of mangrove fishers also increases than case 3_moderate condition
for case 3A and case 3B but does not vary much for these conditions. But the number
of mangrove fishermen who have to pay loan is the lowest for case 3B (highest 515).
Figure 10(C) indicates that the values of catching capacity of household fishermen
and number of household fishermen to pay loan do not change significantly for case
3A and case 3B. From Fig. 10(D), it is noticed that crop production capacity and
the number of farmers who have loan increase and decrease respectively from case
3_moderate for case 3A and case 3B, and case 3 gives the best performance. Again,
from Fig. 10(E), it can be realized that efficiency of FO remains constant for all
conditions where efficiency of UPO increases for case 3A and case 3B than case
3_moderate.
Figure 11 is for showing change of pressure on ecosystem for different livelihood
groups with time comparing with the favorable scenario of case 3 (case 3_favorable).
Figure 11(A) shows that in case 3A and case 3B bawalis give a very low pressure
(0.07 to 0.08) on ecosystem than case 3_favorable (0.14) most of the time, and this
value does not change significantly for case 3A and case 3B. From Fig. 11(B), it
is noticed that in case 3_favorable condition, mangrove fishermen create pressure
of almost 0.23 whereas for case 3A and case 3B, this value remains as 0.2 most
of the times. Figure 11(C) represents that the pressure of household fishers on the
ecosystem has almost similar values for case 3_favorable and case 3B (0.42 and 0.4
respectively) where the value is comparatively lower in case 3A about 0.35) most
of the times. From Fig. 11(D), it can be realized that for case 3A farmers give the
lowest pressure on ecosystem (about 0.55) and this value is significantly higher for
case 3B (about 0.55) and case 3_favorable (about 0.6) most of the times.
So, by analyzing Figs. 10 and 11, it can be decided that in the moderate condition
of case 3 we need to keep conservation_growth_rate between 137.5 and 150 Metric
ton, movement_cost of mangrove fishermen between 0.25 and 0.3125 Metric ton,
production_cost_fish for household fisher between 0.5 and 0.25 Metric ton, and
fertilizer_cost of farmers between 0.625 and 0.75 Metric ton, in order to maintain
better livelihood security with ecosystem sustainability.
4.2 Livelihood and Ecosystem Sustainability
NetLogo simulation results indicate that the livelihood activities of the liveli-
hood groups can be considerably influenced, both favorably and adversely, by the
ecosystem characteristics and the institutional efficiencies. These influences vary
in a range of cases and scenarios represented by the agents and global variables.
Sustainability of the livelihood activities is also indicated by a change in the number
of livelihood groups able to continue their activities in the long term. In general,
the extraction or production capacities is relatively high in the favorable scenarios
Fig. 11 Pressure on the (A) Bawalis

ecosystem in different
conditions caused by: (A)
Bawalis; (B) Mangrove
fishers; (C) Household
fishers; and (D) Farmers
(B) Mangrove fishers
(C) Household fishers
(D) Farmers
where the livelihood activities can sustain over a long period of time. However, this
can lead to increased pressure on the ecosystem. In the critical scenario, low liveli-
hood activity is associated with low pressure on the ecosystem, but reduced capacity,
reflected in increase in number of people taking loans. Hence, this does not represent
an optimal case, where livelihood is maximized and at the same time ecosystem is
best preserved. The optimal balance may be found under the moderate scenario.
Ecosystem sustainability is, however, more sensitive to and dependent on the insti-
tutional policy for conservation of the ecosystem and preventing overexploitation.
Providing financial security to the vulnerable livelihood groups encourages a lower
level of resource exploitation thus indirectly contributing to ecosystem sustainability.
4.3 SES Equilibrium and Tipping Points
Livelihoods of the main communities considered in this study—Farmers, Bawalis,

and Fishers—are closely related. They are part of the same SES and share resources,
and often shift their livelihoods either to overcome a temporary financial crisis or on
a more regular seasonal basis. This situation leads to a more dynamic complexity
than what could be represented in this simplified model. However, it is evident that
the SES reaches an equilibrium condition and averts a tipping point if institutional
policies are formulated to support optimum levels of livelihood activities and to
ensure that the ecosystem resources do not deplete in the long term.
4.4 Role of Institutional Agents: Sensitivity of Policy Change
As evident from the results, the institutional agents have significant influence on both
the livelihood activities and ecosystem response. Efficient performance of the institu-
tions is therefore essential to conserve the mangrove and secure the livelihoods. The
ABM results indicate that protection and conservation of golpata in the Sundarbans
is not only good for the livelihood groups but it also sustains the overall ecosystem.
Similarly, fish conservation in the mangroves ensures continued ecosystem services
as well as higher levels of fish catch.
We observed that comprehensive institutional policies to ensure mangrove SES
sustainability and participatory management are missing in Bangladesh. All current
policies are arbitrarily set either for a specific livelihood group, or within an adminis-
trative boundary of the local government, or only for forest management. The conser-
vation rules and extraction licensing are based on approximate estimates rather than
scientific observations such as fish catch and composition assessment or satellite
data-based assessment.
5 Conclusion
ABM simulations and stakeholder interactions indicate that the vulnerable livelihood
groups of the Sundarbans mangrove SES require policy support to sustain their
livelihoods. Economic returns at the household level increase with an improved
level of institutional decision-making based on scientific assessment. This, in turn,
sustains the mangrove-dependent livelihoods, optimizes the availability of ecosystem
resources, and ensures continued ecosystem services.
We conclude that a comprehensive policy is essential to ensure sustainability
of the mangrove SES. Such policy should outline the roles of relevant institutions
and stakeholders, the approach to science-based decision-making, the method to
identify vulnerabilities of the livelihood groups and the ecosystem, and the priorities
to provide support to the livelihood groups.
Acknowledgements We acknowledge the funding received from the UKRI GCRF Living Deltas
Hub under Grant Reference NE/S008926/1.
References
Acosta, L.A., D.B. Magcale-Macandog, J. Manuta, A.J. Jacildo, E.R. Abucay, J.P. Talubo, E.A.
Eugenio, P.B.M. Macandog, I.M.H. Escamos, and M. Gonzalo. 2018. Agent-Based Modeling
of Sustainable Livelihoods for Vulnerable People to Adapt to the Impacts of Flood-Related
Landslides in the Philippines. Technical Report, The Oscar M. Lopez Center for Climate Change
Adaptation and Disaster Risk Management Foundation, Inc. and University of the Philippines
Los Baños.
An, Li. 2012. Modeling Human Decisions in Coupled Human and Natural Systems: Review of
Agent-Based Models. Ecological Modelling 229: 25–36. https://doi.org/10.1016/j.ecolmodel.
2011.07.010.
An, Li, Alex Zvoleff, Jianguo Liu, and William Axinn. 2014. Agent-Based Modeling in Coupled
Human and Natural Systems (CHANS): Lessons from a Comparative Analysis. Annals of the
Association of American Geographers 104 (4): 723–745. https://doi.org/10.1080/00045608.2014.
910085.
Banglapedia. 2021. The Sundarbans. https://en.banglapedia.org/index.php/Sundarbans,_The#:~:
text=Sundarbans%2C%20The%20largest%20single%20block,with%20the%20bay%20of%20b
engal.&text=The%20Sundarbans%20was%20originally%20measured,of%20about%2016%
2C700%20sq%20km.
Castle, Christian J.E., and Andrew T. Crooks. 2006. Principles and Concepts of Agent-
Based Modelling for Developing Geospatial Simulations. Working Papers Series, Paper 110,
September2006, Centre for Advanced Spatial Analysis, University College London.
Cole, Daniel H., Graham Epstein, and Michael D. McGinnis. 2019. Combining the IAD and
SES Frameworks. International Journal of the Commons 13 (1): 1–32. https://doi.org/10.18352/
ijc.xxx.
Dobbie, Samantha, Kate Schreckenberg, James G. Dyke, Marije Schaafsma, and Stefano Balbi.
2018. Agent-Based Modelling to Assess Community Food Security and Sustainable Livelihoods.
Journal of Artificial Societies and Social Simulation 21 (1): 1. https://doi.org/10.18564/jasss.3639.
Forrester, John, Richard Greaves, Howard Noble, and Richard Taylor. 2014. Modeling Social-
Ecological Problems in Coastal Ecosystems: A Case Study. Wiley Periodicals 19(6) Complexity
73. https://doi.org/10.1002/cplx.21524.
Getzner, Michael and Muhammad Shariful Islam. 2013. Natural Resources, Livelihoods, and
Reserve Management: A Case Study from Sundarbans Mangrove Forests, Bangladesh. Interna-
tional Journal of Sustainable Development and Planning. https://doi.org/10.2495/SDP-V8-N1-
75-87.
Habib, Kazi Ahsan, Amit Kumer Neogi, Najmun Nahar, Jina Oh, Youn-Ho Lee and Choong-Gon
Kim. 2020. An Overview of Fishes of the Sundarbans, Bangladesh and Their Present Conservation
Status. Journal of Threatened Taxa 12 (1): 15154–15172. https://doi.org/10.11609/jott.4893.12.
1.15154-15172.
Hoq, M. Enamul. 2007. An Analysis of Fisheries Exploitation and Management Practices in Sundar-
bans Mangrove Ecosystem Bangladesh. Ocean & Coastal Management 50 (5–6): 411–427.
https://doi.org/10.1016/j.ocecoaman.2006.11.001.
Hossain, M. Sarwar, John Dearing, Felix Eigenbrod, and Fiifi Amoako Johnson. 2017. Opera-
tionalizing Safe Operating Space for Regional Social-Ecological Systems. Science of the Total
Environment 584–585: 673–682. https://doi.org/10.1016/j.scitotenv.2017.01.095.
http://www.gaburaup.satkhira.gov.bd. 2021.
Kabir, K.A., S.B. Saha, and Michael Phillips. 2019. Aquaculture and Fisheries in the Sundar-
bans and Adjacent Areas in Bangladesh: Resources, Productivity, Challenges and Opportunities.
Aquaculture and Fisheries in the Sundarbans and Adjacent Areas in Bangladesh: Resources,
Productivity, Challenges and Opportunities. In Sen H. (ed), The Sundarbans: A Disaster-Prone
Eco-Region. Coastal Research Library, vol 30. Springer, Cham. https://doi.org/10.1007/978-3-
030-00680-8_9.
Lippe, M., M. Bithell, N. Gotts, et al. 2019. Using Agent-Based Modelling to Simulate Social-
Ecological Systems Across Scales. GeoInformatica 23: 269–298. https://doi.org/10.1007/s10
707-018-00337-8.
Mallick, Bishawjit Rupkatha, Jude N. Priodarshini, Bangkim Biswas Kimengsi, Alexander E.
Hausmann, Safiqul Islam, Saleemul Huq, and Joachim Vogt. 2021. Livelihoods Dependence
on Mangrove Ecosystems: Empirical Evidence from the Sundarbans. Current Research in
Environmental Sustainability 3 (2021). https://doi.org/10.1016/j.crsust.2021.100077.
Macal, Charles M., and Michael J. North. 2009. Agent-Based Modeling and Simulation. Proceedings
of the 2009 Winter Simulation Conference, M. D. Rossetti, R. R. Hill, B. Johansson, A. Dunkin
and R. G. Ingalls, eds., December 13–16, 2009, Austin, Texas, USA, pp. 86–98. https://doi.org/
10.1109/WSC.2009.5429324.
Martin, R., and M. Schlüter. 2015. Combining System Dynamics and Agent-Based Modeling to
Analyze Social-Ecological Interactions—An Example from Modeling Restoration of a Shallow
Lake. Frontiers in Environmental Science 3: 66. https://doi.org/10.3389/fenvs.2015.00066.
Miyasaka, T., Q.B. Le, T. Okuro, et al. 2017. Agent-Based Modeling of Complex Social–Ecolog-
ical Feedback Loops to Assess Multi-Dimensional Trade-Offs in Dryland Ecosystem Services.
Landscape Ecol 32: 707–727. https://doi.org/10.1007/s10980-017-0495-x.
Mozumder, Mohammad Mojibul Hoque, Md. Mostafa Shamsuzzaman, Md. Rashed-Un-Nabi,
Ehsanul Karim. 2018. Social-Ecological Dynamics of the Small-Scale Fisheries in Sundarban
Mangrove Forest, Bangladesh. Aquaculture and Fisheries. https://doi.org/10.1016/j.aaf.2017.
12.002.
Ostrom, Elinor. 1990. Governing the Commons. Cambridge, UK: Cambridge University Press.
Ostrom, Elinor, M. Janssen, and J.M. Anderies. 2007. Going Beyond Panaceas. Proceedings of
the National Academy of Sciences 104 (39): 15176–15178. https://doi.org/10.1073/pnas.070228
8104.
Rounsevell, M.D.A., D.T. Robinson, and D. Murray-Rust. 2012. From Actors to Agents in Socio-
Ecological Systems Models. Philosophical Transactions of the Royal Society B367: 259–269.
https://doi.org/10.1098/rstb.2011.0187.
Sarker, Khalekuzzaman. 2011. Fisher Livelihoods in the Sundarbans. Rural Livelihoods and
Protected Landscape: Co-management in the Wetlands and Forests of Bangladesh.
Schlüter, M., R.R.J. Mcallister, R. Arlinghaus, N. Bunnefeld, K. Eisenack, F. Hölker, E.J. Milner-
Gulland, B. Müller, E. Nicholson, M. Quaas, and M. Stöven. 2012. New Horizons for Managing
the Environment: A Review of Coupled Social-Ecological Systems Modeling. Natural Resource
Modeling 25 (1): 219–272. https://doi.org/10.1111/j.1939-7445.2011.00108.x.
Schlüter, M., J. Hinkel, P.W.G. Bots, and R. Arlinghaus. 2014. Application of the SES Framework
for Model-Based Analysis of the Dynamics of Social-Ecological Systems. Ecology and Society
19 (1): 36. https://doi.org/10.5751/ES-05782-190136.
Schulze, Jule, Birgit Müller, Jürgen. Groeneveld, and Volker Grimm. 2017. Agent-Based Modelling
of Social-Ecological Systems: Achievements, Challenges, and a Way Forward. Journal of
Artificial Societies and Social Simulation 20 (2): 8. https://doi.org/10.18564/jasss.3423.
Tisue, Seth and Uri Wilensky. 2004. NetLogo: A Simple Environment for Modeling Complexity.
International Conference on Complex Systems, Boston, May 16–21, 2004.
Training_Report_of_Sundarbans_Nypa_Collector 2012. https://www.ccec-bd.org › files › reports.
Virapongse, Arika, Samantha Brooks, Elizabeth Covelli Metcalf, Morgan Zedalis, Jim Gosz,
Andrew Kliskey, and Lilian Alessa. 2016. A Social-Ecological Systems Approach for Envi-
ronmental Management. Journal of Environmental Management 178: 83–91. https://doi.org/10.
1016/j.jenvman.2016.02.028.
Voinov, Alexey, Karen Jenni, Steven Gray, Nagesh Kolagani, Pierre D. Glynn, Pierre Bommel,
Christina Prell, Moira Zellner, Michael Paolisso, Rebecca Jordan, Eleanor Sterling, Laura Schmitt
Olabisi, Philippe J. Giabbanelli, Zhanli Sun, Christophe Le Page, Todd K. SondossElsawah, Klaus
Hubacek BenDor, Bethany K. Laursen, Laura Basco-Carrera. AntonieJetter, Alison Singer, Laura
Young, Jessica Brunacini, and Alex Smajgl. 2018. Tools and Methods in Participatory Modeling:
Selecting the Right Tool for the Job. Environmental Modelling & Software 109: 232–255. https://
doi.org/10.1016/j.envsoft.2018.08.028.
Wikipedia. 2021. Sundarbans. https://en.wikipedia.org/wiki/Sundarbans.
Yan, Huimin, Lihu Pan, Zhichao Xue, Lin Zhen, Xuehong Bai, Yunfeng Hu, and He-Qing Huang.
2019. Agent-Based Modeling of Sustainable Ecological Consumption for Grasslands: A Case
Study of Inner Mongolia, China. Sustainability 11 (8): 2261. https://doi.org/10.3390/su11082261.
Drought Management by Integrated
Approaches in T. Aman Rice Season
to Escalate Rice Productivity in Drought
Prone Regions of Bangladesh
Debjit Roy, Md. Belal Hossain, Mohammad Rezoan Bin Hafiz Pranto,
and Md. Towfiqul Islam
Abstract Erratic rainfall often causes water shortage for wet season rice (T. Aman)
cultivation, which has a negative impact on crop growth and yield. The long-term
rainfall analysis of drought prone North-West region of Bangladesh indicated that
moderate to severe drought occurred in this region if monsoon ceased between the last
week of September and the first week of October. Integrated management approaches
could help reduce the impact of drought on T. Aman rice production. The approaches
include (i) applying supplemental irrigation when needed after rainfall stops; (ii)
introducing drought-escaping, short duration rice variety; and (iii) adjustment of
transplanting time to escape drought spells. Applying supplemental irrigation two
to three times during the critical stages of rice (from flowering to grain filling),
reduced 20–40% yield loss over fully rainfed conditions. Timely transplantation of
drought-escaping, short duration variety allows the crop to mature before a drought
is exposed. Farmer has comparative flexibility when choosing the transplanting date
for the long duration variety. If T. Aman rice could be transplanted between 10 and
24 July, drought implications on critical stages could be avoided. Integrated drought
management approaches have been found to increase the productivity of T. Aman
rice in the drought suffered regions.
Keywords Drought management · Erratic rainfall · T. Aman rice · Supplemental

irrigation · Transplanting window
1 Introduction
Rainfed Aman (T. Aman) rice is one of the major crops representing the Bangladesh
agricultural sector, and it constitutes about 39% of total rice production of Bangladesh
(BBS 2019). In the last few decades, Bangladesh has achieved tremendous advance-
ments in agricultural technology such as developing high yielding rice varieties and
D. Roy (B) · Md. B. Hossain · M. R. B. H. Pranto · Md. T. Islam

Irrigation and Water Management Division, Bangladesh Rice Research Institute, Gazipur,
Bangladesh
352 D. Roy et al.
introducing modern irrigation water management practices. Despite the technolog-

ical progress, the fate of T. Aman rice production still largely depends on timely
occurrence of rainfall. The rainfall amount and distribution pattern throughout the
season determine the irrigation demand and act as a prime climatic factor that affects
rice grain yield (Sattar and Parvin 2009a). The changed climatic situation is causing
erratic rainfall distribution, reduces the rainfall occurrences at critical period of crop,
thus results water stress to crop due to insufficient moisture in soil profile (Rosen-
zweig et al. 2002). In the coming future, lengthy rainless dry spells will bring agri-
culture droughts and raise the water demand for rainfed rice (Fischer et al. 2007).
Agricultural drought acts as the most influential abiotic stress hampering T. Aman
yield (Biswas 2011). Agricultural drought usually occurs when there is a lack of
adequate rainfall. Due to insufficient rainfall, soil moisture depletes and crop root
zone suffers from water deficit. When water storage in vadose zone becomes limited,
the soil profile cannot provide enough water for the crop to sustain. As such, the situ-
ation is termed as agricultural drought, which significantly affects rice plant growth
and yield (Biswas et al. 2019, MDMR-CEGIS 2013).
Annually, on average, the north-west and central parts of Bangladesh receive
1500 mm rainfall, when the north-east and south-east parts get more than 3000 mm
rainfall (Roy et al. 2014). Around 90% of total annual rainfall occurs between April
and October, commonly known as the monsoon season. T. Aman rice is typically
cultivated between these months in a year. However, in some years, rainfall caused
by monsoon ceases early in between mid-September and early October. This creates
irrigation demand for existing T. Aman rice being grown in the field and causes
moderate to severe agricultural drought (Saleh 1991). Agricultural drought occur-
rence in September and October months often leads to drastic reduction of rice yield
because the crop is usually in between its flowering and grain filling stages of growth,
which are very sensitive to water stress (Sattar and Parvin 2009b). T. Aman rice is
very sensitive to water stress at this reproductive stage because water shortage at
flowering stage reduces crop yield significantly (Yang et al. 2019). The loss in yield
is usually irreversible since rice plant’s physiological features are completely depen-
dent on the fulfilment of water needs for its reproductive phase. Drought occurrence
at the vegetative phase causes 20–25% yield loss, whereas the yield loss could be
more than 50% if drought occurs at the reproductive phase (Yang et al. 2019; Sattar
1993). Drought management to mitigate the hike in crop water requirements during
this period can ensure sustainable T. Aman production (BRRI 1991; Haq et al. 1985;
Islam 2007; Khan 1979; Roy et al. 2010; Saleh 1987).
Many researchers (Ibrahim 2001, Islam et al. 2009, Biswas et. al. 2019, Islam and
Biswas 2010) have suggested several options for drought management. However, a
complete management package is still lacking for farmers in drought prone areas
where they can only choose one appropriate drought escape or mitigation technique
depending on their resources. The present study aimed to evaluate the opportunity
to apply integrated drought management approaches consisting of three practices to
avoid the drought risk in T. Aman rice cultivation. These are: (i) applying supple-
mental irrigation when needed after rainfall stops; (ii) introducing drought-escaping,
short duration rice variety; and (iii) adjustment of transplanting time to escape drought
Drought Management by Integrated Approaches in T. Aman Rice … 353
spells. Integrated drought management approaches would be effective in moderate

to severe drought affected areas and it will ensure that T. Aman yield would not
decrease due to water stress during critical stages of crop production.
2 Methodology
2.1 Study Area
The North-West region of Bangladesh usually suffers from very severe to moderate
agricultural drought (Fig. 1), mainly affecting the production of T. Aman rice. For
the experimentation of different approaches for drought management, the trials were
conducted in three locations in the North-West region, namely Kushtia (23.92° N,
89.2° E), Pabna (24.35° N, 89.73° E), and Rajshahi (24.72° N, 88.97° E). The exper-
imental sites belong to the Agro-Ecological Zone (AEZ)-11, i.e. the High Ganges
River flood plain. The region has medium–high, typical rice growing lands. The soil
is moderately fertile and varies from clay loam to sandy loam texture. The average
high and low temperatures are 37.8° C and 9.2° C, 31.2° C and 20.8 °C, and 32.2° C
and 20.6° C in Kushtia, Pabna and Rajshahi, respectively. The mean annual rainfall
is 1478 mm, 1603 mm, and 1542 mm in Kushtia, Pabna, and Rajshahi, respectively.
The long-term monthly rainfall distribution in these locations, as presented in Fig. 2,
reveals that the premature cessation of monsoon after September is increasing in
frequency and becoming the normal rainfall pattern in the North-West region.
2.2 Supplemental Irrigation
The first study was conducted for three treatments with four replications. Among the
treatments, first treatment was rainfed, i.e. no supplemental irrigation was applied
throughout the growth duration. For the second treatment, two supplemental irriga-
tions were applied at critical stages of T. Aman rice (flowering to grain filling stage),
and for the third treatment, three supplemental irrigations were applied at critical
stages. The second study was set up with five treatments and replicated thrice. The
treatments were: T1—solely rainfed, T2—supplemental irrigation applied just after
transplanting, T3—supplemental irrigation applied from transplanting to panicle
initiation, T4—supplemental irrigation applied during transplanting to flowering, and
T5—supplemental irrigation applied when perched water table had gone 15 cm below
soil surface. The perched water table was measured by installing a polyvinyl chloride
(PVC) pipe. BR11 was transplanted in both experiments and fertilizer management
practices recommended by the Bangladesh Rice Research Institute (BRRI) were
followed. Seedlings of 30 days were transplanted at 20 × 20 cm spacing after proper
354 D. Roy et al.
Fig. 1 T. Aman rice growing area of Bangladesh under different agricultural drought conditions.
The study area (inside the box) in North-West region is under moderate to very severe drought
Fig. 2 Monthly distribution of normalize (1981–2017) effective rainfall (ER) and potential
evapotranspiration (ET0 ) in Kushtia, Pabna and Rajshahi areas
land preparation. Individual plot size was 4 × 5 m with 20 × 40 cm levees. Each

plot was separated by buffer zone of 1 × 1 m.
2.3 Short Duration T. Aman Rice Variety Cultivation
The experiment was set up with four treatments and was replicated thrice. The treat-
ments were: T1 = Date of Transplanting 10 July; T2 = Date of Transplanting 17
July; T3 = Date of Transplanting 24 July; and T4 = Date of Transplanting 31 July.
A well-recommended short duration variety BRRI dhan33, especially suitable for
the North-West region of Bangladesh, was used for the experiment during T. Aman
season. The drought amount was calculated using a water balance-based drought
model. Cultural and fertilizer management practices recommended by BRRI were
followed in growing the crop. Seedlings were raised outside the experimental field
and 30 days old seedlings were transplanted after proper land preparation with 20 ×
20 cm spacing. Individual plot size was 4 × 5 m, separated by 20 × 40 cm levees.
2.4 Adjustment of Transplanting Time
In this study, rain gauges were installed near the experimental field to collect rain-
fall data throughout the growing period. The daily weather data of three North-
West region districts, namely Kushtia, Pabna, and Rajshahi, were collected from
the Bangladesh Meteorological Department (BMD). FAO developed CROPWAT 8.0
356 D. Roy et al.
model was used to determine potential evapotranspiration, actual crop water require-
ment, effective rainfall, and irrigation demand (ID) from collected daily weather
data. Six transplanting dates, namely 10 July, 17 July, 24 July, 31 July, 7 August,
and 14 August, were considered for this experiment. A long duration T. Aman rice
variety, BR11, was taken as test cultivar for this experiment. The seasonal and phase-
wise crop water requirement, effective rainfall, and irrigation demand for Kushtia,
Pabna and Rajshahi, under varying transplanting dates during T. Aman, 2013 to T.
Aman, 2015, were calculated. A threshold yield of 5.5 t/ha equal to the national yield
average (BRRI 2019) was considered as selection criteria for identifying suitable
transplanting window.
3 Results
3.1 Supplemental Irrigation
Table 1 presents the treatment and replication-wise water requirement, water applied,
yield obtained, and yield increment due to the application of supplemental irrigation
at critical stages for the first study of supplemental irrigation application. The research
findings showed that two supplemental irrigations, during the critical stages of T.
Aman rice (i.e. flowering to grain filling), could reduce yield loss by 18–20% over
completely rainfed conditions. The yield loss would be minimized more, by 40–45%,
if three supplemental irrigations were applied.
Table 1 Effect of supplemental irrigation on T. Aman rice yield

Treatment Replication Water Water Yield (t/ha) Yield Average
requirement applied increased yield
(mm) (mm) (%) increased
(%)
Rainfed R1 833 833 3.7 -
R2 840 840 3.5 -
R3 892 892 3.55 -
R4 957 957 3.72 -
Two R1 810 910 4.56 23.2 22.0
irrigations R2 779 879 4.4 18.9
at critical
stages R3 860 960 4.38 18.4
R4 884 984 4.71 27.3
Three R1 658 808 5.29 43.0 41.9
irrigations R2 689 839 5.17 39.7
at critical
stages R3 747 897 5.35 44.6
R4 762 912 5.19 40.3
Fig. 3 Rainwater vs. Water requirement for different treatments
From the second study of supplemental irrigation, it was observed that the period
from panicle initiation (PI) to flowering was critical, because difference between the
water requirement (560 mm) and rainfall received (227 mm) during this period was
very high compared to other periods, as depicted in Fig. 3. This difference caused
maximum amount of water shortage during this period, as rainfall was not sufficient to
meet the crop water requirement. Hence, at this period, when supplemental irrigations
were applied in treatment T4 and T5 in addition to rainfall in order to meet the water
stress, the crop did better than that of other treatments.
3.2 Short Duration T. Aman Variety
In this experiment, after transplanting the short duration T. Aman rice variety BRRI
dhan33, the highest total drought amount was observed for transplanting date of
31 July and the lowest total for transplanting date of 17 July (Table 2). T. Aman
rice transplanted on 10 July and 17 July faced drought during the vegetative stage, of
Table 2 Drought amount at different growth stages of BRRI dhan33

Treatment Vegetative Stage (mm) Reproductive Stage Ripening Stage (mm) Total
(mm)
10 July 61.97 0 0 61.97
17 July 20.89 0 16.39 37.28
24 July 0 0 55.99 55.99
31 July 0 1.55 79.34 80.89
358 D. Roy et al.
61.97 mm, and 20.89 mm, respectively. Drought at ripening stage was observed for all
dates of transplanting except 10 July, and the highest was found for 31 July. Drought
amount during the reproductive stage was not considerable in any treatment. Figure 4
represents transplanting date-wise drought pattern throughout the growth duration of
BRRI dhan33 and demonstrates the safe escaping of drought at reproductive stages.
BRRI dhan33, that was transplanted on 24 July, yielded higher (4.60 t/ha) than the
other treatments for that variety (Table 3). In this case, the rice did not suffer from
drought during its vegetative and reproductive stages. However, it faced relatively
higher drought amount in the ripening stage. The next highest yield was observed
for date of transplanting 31 July (4.23 t/ha). In the other two treatments, yields were
comparatively lower than that of 24 July and 31 July transplanting. Considerable
amount of drought was observed in the vegetative stages of rice transplanted on 10
July and 17 July. Drought at the ripening stage did not have a negative impact on
rice yield at any treatment. Therefore, considering the drought amount at different
growth stages, short duration T. Aman rice variety like BRRI dhan33 could easily
escape drought during the critical stages.
Fig. 4 Drought patterns for BRRI dhan33
Table 3 Yield performance

Treatment Plant height (cm) Panicle/m2 Yield (t/ha)
of BRRI dhan33
10 July 108 231 3.85
17 July 105 228 3.93
24 July 106 266 4.60
31 July 109 241 4.23
3.3 Adjustment of Transplanting Time
From Table 4, it can be observed that the vegetative phase did not have any irrigation
demand in all three years (2013–2015). In all years, irrigation demand at the repro-
ductive stage showed increasing trend due to transplanting delay. In 2014, irrigation
demand was the highest in the reproductive stage due to limited rainfall failing to
effectively meet the water requirement for that stage. However, irrigation demand in
the reproductive and ripening stages was considerable in all three years except 2013.
Table 4 Growth stage-wise effective rainfall (ER in mm) and Irrigation demand (ID in mm) for
different transplanting dates during T. Aman 2013 to 2015 at Kushtia
Growth Phase 10-Jul 17-Jul 24-Jul 31-Jul 07-Aug 31-Aug
2013
Vegetative ER 358 350 349 344 335 320
phase ID 0 0 0 0 0 0
Reproductive ER 118 115 111 84 80 55
phase ID 0 0 0 18 21 44
Ripening phase ER 48 29 0 0 0 0
ID 50 71 99 97 98 92
Seasonal ER 524 494 461 428 415 386
ID 0 3 22 35 30 22
2014
Vegetative ER 435 435 391 356 333 317
phase ID 0 0 0 0 0 0
Reproductive ER 78 75 77 58 43 27
phase ID 62 60 52 67 80 89
ID 81 109 106 102 101 91
Seasonal ER 539 510 467 414 376 344
ID 8 22 55 93 105 93
2015
Vegetative ER 467 450 429 395 366 336
phase ID 0 0 0 0 0 0
Reproductive ER 81 81 82 69 59 49
phase ID 45 43 41 52 60 59
ID 71 82 94 89 87 74
Seasonal ER 580 551 512 465 426 386
ID 0 0 0 0 19 8
360 D. Roy et al.
Fig. 5 Relationship between

grain yield and irrigation
demand of BR11 at
reproductive phase in
Kushtia
Figure 5 illustrates yield response to irrigation demand (ID) in the reproduc-

tive phase. From the relationship (P < 0.01), it could be reported that the grain
yield decreased by 0.0138 t/ha if irrigation demand (ID) increased by 1 mm. This
result supports the findings from Yang et al. (2019), which stated that the effect of
drought stress at flowering stage was significant as it affected physiological traits and
decreased grain yield. In a study conducted in Tamil Nadu, India, researchers have
found a reduction of grain yield by 28% due to mild water stress in the reproduc-
tive phase of rainfed rice during 1999–2000 (Babu et al. 2003). Water stress at the
reproductive stage caused significant grain yield reduction (Kamoshita et al. 2008).
BR11 yield performance was analysed for the three locations and has been illus-
trated in Fig. 6. The yields were estimated using the relationship developed between
grain yield and irrigation demand (ID). For Kushtia, the yields of BR11 were equal to
or higher than the threshold limit (5.5 t/ha) for transplanting window of 10–24 July.
In Pabna, considering the threshold limit, 10–24 July was found as the suitable trans-
planting window (Fig. 6). In Rajshahi, the yields for different transplanting dates did
not come close to the threshold limit. Hence, considering the reduced limit as 5 t/ha,
the suitable transplanting window was found to be 10–24 July. In all locations, yield
went down drastically if the transplanting date was outside of the suitable window.
4 Discussion
Supplemental irrigation plays a vital role in alleviating the impact of drought, espe-
cially on T. Aman rice production. T. Aman rice always suffers from either short
or long-spell drought at the reproductive stage of the crop. Sattar (1993) stated that
water stress at the vegetative stage can cause a yield loss of about 25% and this
may be as high as 50% at the reproductive stage. Therefore, supplemental irrigation
should be applied by any means from any source. It has also been observed from the
effect of drought on Aman rice growth phases (BRRI 2008), that under a drought
spell, percentage of yield reduction was higher at the reproductive phase than at the
Fig. 6 Yield performance of BR 11 over threshold under different transplanting dates at (a) Kushtia,
(b) Pabna and (c) Rajshahi
ripening phase. Hence, the reproductive phase is considered more vulnerable to water
stress in comparison with the ripening stage. BRRI research results showed that two
or three supplemental irrigations in proper time could increase T. Aman rice produc-
tion by 40–45% (Table 1). However, Incremental yield was completely dependent
on water shortage minimization by supplemental irrigation in excess to rainfall. The
period from panicle initiation to flowering stages was critical as maximum water
362 D. Roy et al.
stress occurred during this period due to insufficient rainfall. T. Aman rice culti-
vation largely depends on rain for its water supply. So, the source of supplemental
irrigation water is a major concern during T. Aman season. Rainwater harvesting by
mini-pond and levee management could be the best solution to this problem. The
stored water in mini-ponds can be used for land preparation and for irrigation under
a drought spell during the vegetative and reproductive stages of T. Aman rice (BRRI
1991). A 15 cm-levee height management can give the highest yield of T. Aman rice
by conservation of maximum effective rainwater. These can be effectively used for
improving land productivity in the rainfed areas.
Cultivation of short duration variety is another effective means for mitigating the
effects of agricultural drought by escaping it altogether. Findings of this study show
that if the cultivar could be transplanted at a recommended time, critical stages of
growth duration like flowering and grain filling stages would be completed before the
drought appeared. However, the yield of short duration rice variety was comparatively
lower than that of the long duration variety. But at least, drought could not be able
to hamper the rice production of short duration variety compared to long duration
variety.
For transplanting long duration variety like BR11, the transplanting window of
10–24 July was suitable for Kushtia and Rajshahi while 10–17 July window was
appropriate for Pabna. The farmers who are interested to cultivate long duration
variety must adopt the recommended transplanting dates so that they do not need to
cultivate short duration variety or apply supplemental irrigation.
Although all three approaches are beneficial in varying extents, farmers are not
always interested or able to pick any of those, and there are a number of factors that
they need to take into consideration. Supplemental irrigation application is usually the
best option to mitigate agricultural drought at latter part of the season. But application
of supplemental irrigation includes cost and labour. To reap maximum benefit from
short duration variety, timely transplanting is critical. However, it may not be possible
to prepare the land for transplanting in due time. Adjusting the transplanting dates
for long duration variety would be a flexible choice for farmers. However, timely
land preparation, availability of seeds and fertilizers, and cost of inputs are all key
factors for transplanting T. Aman rice within the suitable window.
5 Conclusion
Drought management using integrated approaches in T. Aman season offers an effec-

tive solution for the North-West region of Bangladesh. The current rainfall pattern
bears evidence that the monsoon is beginning to cease at the end of September, prop-
agating the permanent risk of drought occurrence at the later part of T. Aman rice
season. Supplemental irrigation is the first suggested approach to mitigate terminal
drought at the critical reproductive stages by applying lifesaving water to the plant.
As a second choice, farmers can cultivate short duration crop variety and escape
the effects of drought during critical stages. Adjusting the transplanting dates for
long duration T. Aman rice cultivar can also successfully overcome the irrigation
demand at the reproductive phase. Delay in transplanting demands more irrigation
and reduces yield.
Acknowledgements The author would like to express heartiest gratitude to Bangladesh Rice
Research Institute (BRRI) for providing research facilities and funding. The authors are also thankful
to Agricultural Land and Water Resources Management (ALAWRM) Research Group, Irrigation
and Water Management (IWM) Division, Bangladesh Rice Research Institute (BRRI) for their kind
support and help throughout the study. The authors would like to convey their special thanks to
ICWFM 2021 organizer and the scientific committee for accepting the abstract, giving the oppor-
tunity to make an oral presentation, and finally selecting the paper for publishing in conference
proceedings.
References
Babu, R., B. Nguyen, V. Chamarerk, P. Shanmugasundaram, P. Chezhian, P. Jeyaprakash, et al.

2003. Genetic Analysis of Drought Resistance in Rice by Molecular Markers. Crop Science 43
(4): 1457–1469. https://doi.org/10.2135/cropsci2003.1457.
BBS (Bangladesh Bureau of Statistics). 2019. “Yearbook of Agricultural Statistics of Bangladesh”
Planning Division, Ministry of Planning. Dhaka: Government of the People’s Republic of
Bangladesh.
Biswas, J.K. 2011. Climate Adversaries: Coping with Rice Agriculture. J. Asiatic Soc. Bangladesh
37: 359–374.
Biswas, J.K., Kabir S, Rahman S, Nahar K, and Hasanuzzaman M. 2019. Managing Abiotic Stresses
with Rice Agriculture to Achieve Sustainable Food Security: Bangladesh Perspective. In Advances
in Rice Research for Abiotic Stress Tolerance, ed. M. Hasanuzzaman et al., 23–46. Duxford:
Woodhead Publishing, Elsevier Inc. https://doi.org/10.1016/B978-0-12-814332-2.00002-2.
BRRI (Bangladesh Rice Research Institute). 1991. Internal Review Report, 149–155. Gazipur:
Irrigation and Water Management Division.
BRRI (Bangladesh Rice Research Institute). 2008. Annual Report. Irrigation and Water Manage-
ment Division. Gazipur: Bangladesh Rice Research Institute.
BRRI (Bangladesh Rice Research Institute). 2019. Modern rice cultivation. Gazipur: Bangladesh
Rice Research Institute.
Fischer, G., F.N. Tubiello, H. Van Velthuizen, and D.A. Wilberg. 2007. Climate Change Impacts
on Irrigation Water Requirements: Effects of Mitigation, 1990–2080. Technological Forecasting
and Social Change 74 (7): 1083–1107.
Haq, K.A., Islam, J., and Sattar, M.A. 1985. Low Irrigation Strategies for Rice Production. Proc.
of the Seminar on Low Water Demanding Crops. Dhaka: BARC.
Ibrahim, A. 2001. Application of Agroecological Zones Database in Drought Management and
Water Availability Assessment. Environment and GIS support project for water sector planning,
7–11. Dhaka.
Islam, M.T. 2007. “Modeling of Drought for Aman Rice in the Northwest Region of Bangladesh.”
Ph. D. Thesis. Mymensingh: Department of IWM, BAU.
Islam, M.T., and J.K. Biswas. 2010. Natural Hazard Management in Rice Production. J Agric Engin
38: 63–69.
Islam, M.T., L.R. Khan, and K.M. Hassanuzzaman. 2009. Integrated Approach for Mitigating
Drought in Transplanted Aman Rice in Northwest Region of Bangladesh. Bangladesh Rice
Journal 14 (1&2): 169–173.
364 D. Roy et al.
Kamoshita, A., R.C. Babu, N.M. Bhupathi, and S. Fukai. 2008. Phenotypic and Genotypic Anal-
ysis of Drought-Resistance Traits for Development of Rice Cultivars Adapted to Rainfed
Environments. Field Crops Res. 109: 1–23. https://doi.org/10.1016/j.fcr.2008.06.010.
Khan, H.R. 1979. Determination of Transplanting Date and Supplementary Irrigation Requirement
for HYV Monsoon Rice Using Rainfall. Bangladesh Journal of Water Resource Research 1 (1):
76–85.
MDMR-CEGIS (Ministry of Disaster Management and Relief-Center for Environmental and
Geographic Information Services). 2013. Vulnerability to Climate Induced Drought: Scenario &
Impact. Center for Environmental and Geographic Information Services. Comprehensive Disaster
Management Programme (CDMP II), Ministry of Disaster Management and Relief, 134.
Rosenzweig, C.E., F. Tubiello, R. Goldberg, E. Mills, and J. Bloomfield. 2002. Increased Crop
Damage in the U.S. from Excess Precipitation under Climate Change. Global Environment
Change A (12): 197–202.
Roy, D., Ahmmed, M.U., and Mahmud, N.H. 2014. Water Management Technologies for
Sustainable Rice Cultivation Under Changing Climate. J. Agric. Engin 41/AE (1): 3–52.
Roy, D., M.T. Islam, and B.C. Nath. 2010. Impact of Supplemental Irrigation Applied at Different
Growth Stages of Transplanted Aman Rice (Oryza Sativa). The Agriculturists 8 (1): 109–114.
Saleh, A.F.M. 1987. The Effect of Supplementary Irrigation on Rice Yield in Bangladesh.
International Rice Research Newsletter 2 (1): 25–26.
Saleh, A.F.M. 1991. Supplementary Irrigation in Bangladesh: Requirements, Benefits, and
Prospects. Proc. of the Special Technical Session Proceedings, International Commission on
Irrigation and Drainage China 1-A: 96–105.
Sattar, M.A. 1993. Water Management and Technology Adoption for Direct Seeded Rice in an
Irrigation System. Dissertation. Philippines: Central Luzon State University.
Sattar, M.A., and Parvin, M.I. 2009a. “Sustainable T. Aman Rice Production in North-West Region
of Bangladesh for Food Security Under Climate Change Situation.” Proc. of International
Conference on Climate Change Impacts and Adoption Strategies for Bangladesh, 289. Dhaka:
BUET.
Sattar, M.A., and Parvin, M.I. 2009b. Assessment of Vulnerability of Drought and Its Remedial
Measures for Sustainable T. Aman Rice Production in the Selected Locations of Bangladesh.”
Proc. of 2nd International Conference on Water and Flood Management (ICWFM 2009), 109.
Dhaka.
Yang, X., B. Wang, L. Chen, P. Li, and C. Cao. 2019. The Different Influences of Drought Stress at
the Flowering Stage on Rice Physiological Traits, Grain Yield, and Quality. Scientific Reports 9:
3742. https://doi.org/10.1038/s41598-019-40161-0.
Actual Evapotranspiration Estimation
Using Remote Sensing: Comparison
of Sebal and Metric Models
Sumit Kumar Saha, Rubel Ahmmed, and Nasreen Jahan
Abstract Evapotranspiration (ET) is the sum of evaporation from earth’s surface and
transpiration from plants to the atmosphere. Accurate quantification of ET is crucial
for hydrologic modeling, optimizing crop production, drought monitoring, irrigation
management, and overall water resource planning. For monitoring evapotranspira-
tion, remote sensing-based Surface Energy Balance Algorithm for Land (SEBAL)
and Mapping Evapotranspiration (ET) at high Resolution with Internalized Calibra-
tion (METRIC) have been applied extensively. However, the complexity of selecting
hot and cold pixels has made the operational use of these models challenging. In this
study, an automated implementation of these models called Land MOD ET mapper
has been tested in two agricultural sites of Bangladesh using Landsat-4-5 TM images,
DEM and weather data from the Modern-Era Retrospective analysis for Research
and Applications, version 2 (MERRA-2). Merra-2 is a global atmospheric reanal-
ysis produced by the NASA. The evapotranspiration calculated by the SEBAL and
METRIC models were compared with the recorded pan evaporation and the Penman–
Monteith method. Comparison of the results from the SEBAL and METRIC shows
some differences in ET estimation. This is probably due to the differences in calcu-
lation of sensible heat and the assumptions of SEBAL and METRIC in extrapolating
instantaneous ET to the daily ET. This study demonstrates the considerable potential
of SEBAL and METRIC models for estimation of spatiotemporal distribution of ET
from Landsat satellite images and Merra-2 weather data for the agricultural regions
of Bangladesh.
Keywords Evapotranspiration · SEBAL · METRIC · Landsat · MERRA-2
S. K. Saha (B) · R. Ahmmed · N. Jahan

Department of Water Resources Engineering, Bangladesh University of Engineering and
N. Jahan
366 S. K. Saha et al.
1 Introduction
1.1 Background of the Study
Evapotranspiration (ET) refers to the conveyance of water vapor from earth’s surface
to the atmosphere. ET is the sum of evaporation and transpiration from the surface
of the earth to the atmosphere. The flow of water to the air from sources such as
the soil, canopy interception, and water bodies is accounted for by evaporation.
On the other hand, transpiration can be defined as loss of water as vapor through
stomata in the leaves. Evapotranspiration is an important phenomenon within the
land–atmosphere interface that regulates the earth’s energy and water cycles. Both
evaporation and transpiration take place at the same time and are influenced by solar
radiation, air temperature, relative humidity, vapor pressure deficit, and wind speed.
Crop characteristics, environmental factors, and cultivation practices influence the
rate of transpiration. Transpiration rates also vary depending on the type of plant.
Evapotranspiration is a vital nexus between terrestrial water, carbon, and surface
energy exchanges and is a fundamental component of the global water cycle (Zhang
et al. 2016). It is an important process in the water cycle because it is responsible
for 15% of the atmosphere’s water vapor (ScienceDaily 2021). Accurate estimation
of ET is essential for undertaking research on climate change, and for managing
agricultural water requirements, drought forecasting and monitoring, and optimal
water resource development and exploitation, among other things. Bangladesh is an
agriculture-based country where Boro rice cultivation during the dry season depends
almost entirely on groundwater irrigation (Kirby et al. 2016). Climate change is
expected to cause a decrease in soil moisture, an increase in daily ET, and these
will increase the irrigation demand and cause further depletion of groundwater level
(Shahid 2011). For South and East Asia, the total annual water withdrawal is roughly
1981 km3 , which is about 50% of the global aggregate, and agriculture constitutes
around 82% of the total freshwater withdrawal in Asia (FAO 2016). Past studies have
reported significant increasing trend of irrigation water quantity due to expanded
areas in the Northwest region of Bangladesh (Mojid et al. 2021). Consequently, high
water withdrawal and the rapid decline of groundwater resources in Bangladesh due
to excessive irrigation during the dry season have become a severe concern that
requires awareness regarding optimal on-field water requirements (GoB and FAO
2018). In addition to the rapid rise in groundwater-dependent irrigation, wastage of
water by irrigating more than the crop water demand also contributes to this alarming
groundwater depletion. Studies have shown that 21% of the water lifted for irrigation
is left unused and lost in Bangladesh (BRAC and BUET, 2013). Therefore, proper
management of groundwater resources for agricultural uses is extremely necessary to
maintain a sustainable balance between groundwater supply and demand, and thereby
ensure food security in the coming decades for Bangladesh. Therefore, an accu-
rate estimation of evapotranspiration is necessary for calculating irrigation demand
accurately and managing available water resources in Bangladesh.
Actual Evapotranspiration Estimation Using Remote Sensing… 367
ET can be estimated using empirical equation or remote sensing algorithm or

can be measured by traditional methods (Liou and Kar 2014). Lysimeter and eddy
covariance method offer in-situ measurement of ET. But they are costly, time-
consuming, and works only at the local scale for a particular crop. A complemen-
tary relationship-based model developed by Anayah and Kaluarachchi (2014) has
been used to compute actual evapotranspiration in a recent study in the Ganges
Delta (Murshed and Kaluarachchi 2018; Murshed, Rahman, and Kaluarachchi 2019)
where meteorological data have been used. The Penman–Monteith formula is used
frequently to overcome this problem, but it requires several climate data and surface
variables that are not easily available (Stannard 1993). Although MODIS satellite
provides actual ET data based on both climate and remote sensing (RS) variables,
the output showed a discrepancy with the observed ET in some studies (Ruhoff
et al. 2013). So, past research on evapotranspiration estimation in Bangladesh have
mainly studied reference ET instead of actual ET because it requires only a moderate
amount of data in comparison with actual ET calculation (Karim et al. 2020). Under
this perspective, Surface Energy Balance Algorithm for Land (SEBAL) or Mapping
Evapotranspiration (ET) at high Resolution with Internalized Calibration (METRIC)
can therefore be an alternative and reliable method of estimating actual ET at local to
regional scales (Li et al. 2009). Remote sensing-based algorithms are cost-effective
and spatiotemporal variation of ET can be monitored easily. The water stress is also
considered here which is not generally considered in the empirical approach. Rojas
and Sheffield (2013) evaluated daily reference evapotranspiration methods using five
different methods including the ASCE-EWRI Penman–Monteith equation in North-
east Louisiana (Rojas and Sheffield 2013). Reyes-González et al. (2017) compared
actual Evapotranspiration values estimated from the METRIC model with other
Methods in Eastern South Dakota(Reyes-González et al. 2017). Bose et al. (2021)
used SEBAL for operational irrigation advisory in Pakistan (Bose et al. 2021).
Past studies on ET estimation in Bangladesh have employed mostly the Penman–
Monteith equation. Sahid, S., (2011) has shown that the climate change will increase
the ET and irrigation rate. He used FAO Penman model for the computation of
ET (Shahid 2011). Some of the past research studied the crop evapotranspiration
(ETc) based on Penman–Monteith method and predicted future irrigation require-
ments of Boro rice (Acharjee et al. 2017). Kanoua and Merkel (2015) used the
Two-Source Trapezoid Model for Evapotranspiration and the SEBAL model to
estimate the spatiotemporal variation in actual evapotranspiration in Titas Upazila,
Bangladesh (Kanoua and Merkel 2015). SEBAL was used to estimate ET in a study
on the availability of groundwater for crops in the Northwest zone of Bangladesh
(Dey et al. 2017). Hossen et al. (2012) reported surface energy components and
evapotranspiration over a double-cropping paddy field in Bangladesh (Hossen et al.
2012).
The objective of the study is to assess the potential of Surface Energy Balance
Algorithm for Land (SEBAL) and Mapping Evapotranspiration at high Resolu-
tion with Internalized Calibration (METRIC) model in estimating ET for agricul-
tural fields of Bangladesh. An automated implementation of these models called
LandMOD ET mapper has been used herein for this purpose at two agricultural
study sites. The estimated ET from SEBAL and METRIC model was also compared
with the ET estimation obtained from the FAO Penman–Monteith equation and the
observed pan evaporation data from Bangladesh Meteorological Department (BMD)
for both stations.
1.2 Study Area
In this study, evapotranspiration has been estimated for Mymensingh Sadar Upazila
and Bogra Sadar Upazila of Bangladesh. Bangladesh Meteorological Department
(BMD) monitors evaporation only at a limited number of agro-meteorological
stations. Two stations out of those have been randomly chosen for the study. The
Mymensingh site (Fig. 1) is a paddy field located at the Bangladesh Agricultural
University Farmland (24.73°N, 90.42°E, and 18 m above the mean sea level). This
1420 sq. m field has been used exclusively for paddy cultivation for about 40 years.
It provides a sufficient upwind fetch of uniform land cover for measuring mass and
energy fluxes using tower-based eddy covariance systems. The soil type of this field
is dark-gray non-calcareous floodplain (UNDP and FAO, 1988) with a sandy loam
texture. The location experiences a tropical monsoon-type climate, with a hot and
Fig. 1 Mymensingh and Bogra study sites

rainy summer, and dry winter. Annual mean air temperature was 298.66 K, and
annual rainfall was 2055 mm. The usual cropping pattern is rice, with two rice crops
per year. One crop is dry-season rice, locally called “Boro” rice, which is cultivated
from late winter (February) to midsummer (May). The other crop is wet-season rice,
locally called “Aman” rice, which is cultivated from late summer (August) to early
winter (December). The field is mostly irrigated during the Boro rice period, while
it is rain-fed during the Aman rice period.
The Bogra site (Fig. 1) is located around the Bangladesh Meteorological Depart-
ment’s climate station at Bogra (24o 51’N and 89o 22’E). Total Area of Bogra Sadar
is 17734 ha in which 150 ha area is used for Annual crop, 1940 ha for Double
crop, 8700 ha for Triple crop and 300 ha for Quadrated crop. The other 110 ha and
11,200 ha are non-cultivable areas. These two stations have been selected as we have
measured pan evaporation data for these locations. The time period of the study has
been chosen based on the availability of pan evaporation data of ET (Table 1).
1.2.1 Weather Data
All required weather data were collected from Modern-Era Retrospective analysis
for Research and Applications, version 2 (MERRA-2) which is a global atmospheric
reanalysis produced by the NASA. Following variables were collected.
I. Instantaneous Relative humidity (%) at the image time (rh_inst)
II. Instantaneous Air temperature in kelvin at the image time (t_inst)
III. Instantaneous Wind speed in m/s at the image time (u_inst)
IV. Instantaneous Incoming solar radiation in wm−2 at the image time
(solar_inst)
V. Maximum daily air temperature in kelvin (tmax_daily)
VI. Mean daily air temperature in kelvin (tmean_daily)
VII. Minimum daily air temperature in kelvin (tmin_daily)
VIII. Daily mean solar radiation wm−2 (solar_daily)
IX. Daily mean Relative humidity % (rh_daily)
X. Daily mean wind speed m/s (u_daily).
1.2.2 Other Remote Sensing Data
Land surface temperature (LST) has been collected from remote sensing lab of
landsat-5. Emissivity has been collected from MERRA-2. Spatial resolution has
been maintained as 30 m by 30 m.
After calculating the reflectance from the digital numbers of the satellite images,
NDVI is calculated by using Eq. 1:
Rλ(NIR) − Rλ(RED)
NDVI = (1)
Rλ(NIR) + Rλ(RED)
Table 1 Description of the data for Mymensingh and Bogra Sadar Upazilla
SL. No Date of Day of Sun Image Collection Earth Sun Shine
Collection Year Elevation Time Distance (d)
Angle Hour Minute
(Degree)
Mymensingh Sadar Upazilla
01 22 Jan, 2007 22 38.257 10 19 0.98419
02 23 Feb, 54 46.007 10 19 0.98944
2007
03 11 Mar, 70 51.227 10 19 0.98717
2007
04 27 Mar, 86 57.676 10 19 0.99782
2007
05 28 pr, 2007 118 65.068 10 19 1.00679
06 30 May, 150 67.715 10 18 1.01371
2007
07 18 Aug, 230 63.161 10 17 1.01225
2007
08 19-Sep, 262 56.826 10 17 1.00457
2007
09 05 Oct, 278 53.813 10 17 1.00005
2007
Bogra Sadar Upazilla
01 24 Jan, 2011 24 38.0341 10 20 0.98439
02 09 Feb, 40 41.4123 10 20 0.98662
2011
03 25 Feb, 56 45.9611 10 20 0.98989
2011
04 13 Mar, 72 51.1584 10 20 0.99392
2011
05 16 May, 136 66.2923 10 19 1.01011
2011
06 08 Nov, 312 43.6547 10 18 1.01403
2011
where Rλ(NIR) is Reflectance of Near Infrared Band (band 4 of Landsat 5), Rλ(RED)
is Reflectance of Near Red Band (band 3 for Landsat 5). Albedo is a measure of how
much radiation is reflected from a surface. It is a ratio of the reflected radiation to the
radiation reached to the ground. Albedo has been computed following Allen et al.
(2007) using the bands of the landsat images and DEM (Digital Elevation Model) of
the selected sites (Allen et al. 2007).
Surface roughness (z0m) was computed using Eq. 2: (Su and Jacobs 2001)
2.5
NDVI
z0m = 0.005 + 0.5 ∗ (2)
NDVImax
where z0m = Surface roughness in meter, NDVI is normalized difference

vegetation index, NDVImax is maximum value of NDVI.
2 Background of Models
2.1 Sebal
The Surface Energy Balance Algorithm for Land (SEBAL) proposed by (Basti-
aanssen et al. 1998) is one of the most widely used remote sensing-based algorithm
for computing energy balance (Bhattarai and Liu 2019). SEBAL estimates actual
evapotranspiration (ETa ) from remotely sensed land surface temperature, surface
emissivity, NDVI, and some routine ground meteorological observation data, i.e.,
sunshine duration and wind speed. The core assumption in SEBAL is that there is a
linear relation between near-surface vertical temperature gradient and land surface
temperature, and this relationship is determined by two extreme hydrological pixels,
i.e., extreme wet/cold and extreme dry/hot pixels (Wang et al. 2014).
SEBAL model estimates the surface net radiation flux Rn (W•m−2 ), the soil heat
flux G (W•m−2 ), and the sensible heat flux H (W•m−2 ) first and then estimates the
latent heat flux LE (W•m−2 ) as an energy residual by Eq. 3:
LE = Rn − G − H (3)
Rn is computed by Eq. 4:
Rn = (1 − α)Rs + εa σ T4a − εs σ T4s − (1 − εs )εa σ T4s (4)
where Rs is the solar shortwave radiation (W•m−2 ), which is estimated by the

empirical parameterization of (Zillman et al. 1972). εs and εa are the surface emissivity
and atmospheric emissivity, respectively; α is albedo (dimensionless); σ is the Stefan-
Botlzmann constant (5.67 × 10−8 W•m−2 •K−4 ); Ts and Ta are the land surface
temperature (in K) and air temperature (in K) at satellite observation time.
G is calculated as a fraction of Rn . The ratio G/Rn is usually defined as a function
of leaf area index (LAI) or Normalized Difference Vegetation Index (NDVI). In
SEBAL, G is calculated by Eq. 5:
G (Ts − 273.15)
= 0.0032αavg + 0.0062αavg
2
1 − 0.978NDVI4 (5)
Rn α
Satellite Image Weather Data
Vegetaon Surface
Index Surface Albedo
Temperature
Momentum
Net Radiaon Soil Heat Flux
Flux
Sensible Heat at extreme

evaporaon points
Sensible Heat Flux
Actual Evapotranspiraon
Fig. 2 SEBAL algorithm
where αavg is the daily average albedo which is readily available as remotely
sensed images. As the variation in albedo under all clear sky presents a U-shape,
high at the sunrise and sunset and low at the noon, therefore a linear algorithm was
suggested to calculate αavg using instantaneous α, i.e., αavg = 1.02α + 0.01 (Teixeira
et al. 2008).
H is calculated by Eq. 6 following Anderson et al. 2011.
H = ρCp (T0 − Ta ) ra (6)
where ρ is the air density (kg•m−3 ), Cp is the air specific heat at constant pressure
(1004 J•K−1 •kg−1 ), T0 is the aerodynamic temperature (K), ra is the aerodynamic
resistance (m•s−1 ). Figure 2 describes the full process.
2.2 METRIC Model
METRIC (Mapping Evapotranspiration at high Resolution with Internalized Cali-

bration) model is another widely used actual ET estimation model (Allen et al.
2007). It is a remote sensing-based model that computes instantaneous ET values as
a residual of the surface energy balance equation which is Eq. (3).
The key component of METRIC model is self-calibration of H using an itera-

tive process based on manually or automatically selected hot and cold pixels. The
METRIC model works best in flat regions. However, it can be used in mountainous
regions if some corrections are added based on digital elevation model, slope and
aspect data (Allen et al. 2007).
METRIC is based on the commonly used principle that ET can be estimated from
the residual term of the surface energy balance equation. The available energy (Rn
− G) is computed first and then partitioned into H and LE. Rn is computed for clear
sky conditions as following Allen et al. (2007) as given in Eq. 7:
Rn = Gsc cos θ dr τsw (1−α) εo εa σ T4a −εo σ T4s (7)
where GSC (1367 W m − 2) is a solar constant, θ is the solar incidence angle

computed from the sun elevation angle (β) where θ = (90 − β), dr is the inverse
squared relative earth-sun distance, and τsw (dimensionless) is the broad-band atmo-
spheric transmissivity (Allen et al. 2007). εa is the atmospheric emissivity computed
as a function of air temperature and vapor pressure (Wilfried 1975).
In METRIC model, the soil heat flux (G) is computed following Tasumi (2003).
This process computes G from Rn , surface temperature, albedo, and NDVI following
Bastiaanssen et al. (1998) on Eq. 8:

G = (Ts − 273.15) α 0.0038α 0.0074α 2 1 − 0.98NDVI4 Rn (8)
The key part of the METRIC model is the determination of H, which uses the
formulation given in Eq. 9:
ρa cc dT
H= (9)
rah
where ρa is air density, cc is the specific heat of the air, dT is temperature difference,
and rah is aerodynamic resistance (Fig. 3).
2.3 Penman–Monteith Equation
To estimate evapotranspiration from various meteorological factors, a large number

of empirical approaches have been developed. Some of them were derived from
popular Penman equation (Penman 1948), which used a mix of an energy balance
and an aerodynamic formula to calculate evaporation from open water, bare soil, and
grass
(Rn − G) + γ λEa
λE = (10)
+λ
Fig. 3 METRIC algorithm (Numata et al. 2017)
where λE = evaporative latent heat flux (MJ m−2 d−1 ), = slope of the saturated
vapor pressure curve [eo /T, where eo = saturated vapor pressure (kPa) and Tmean =
daily mean temperature (°C)]; Rn = net radiation flux (MJ m−2 d−1 ), G = sensible
heat flux into the soil (MJ m−2 d−1 ), 3 = psychometric constant (kPa °C−1 ), and Ea
= vapor transport of flux (mm/day).
A bulk surface resistance term was introduced in various derivations of the Penman
equation (Monteith 1965), and the resulting equation is now known as the Penman–
Monteith equation, which can be stated for daily values as in Eq. 11:
ρa Cp (eos −ea )
(Rn − G) + 86400 rav
λET0 = (11)
+ γ (1 + rs
rav
)
where ρa = air density (kg m−3 ), Cp = specific heat of dry air, es o = mean
saturated vapor pressure (kPa) computed as the mean eo at the daily minimum and
maximum air temperature (°C), rav = bulk surface aerodynamic resistance for water
vapor (s m−1 ), ea = mean daily ambient vapor pressure (kPa), and rs = the canopy
surface resistance (s m−1 ). An updated equation was recommended by FAO (Allen
et al. 1998) known as the FAO-56 Penman–Monteith equation, simplifying Eq. 11
by utilizing some assumed constant parameters for reference crop. It was assumed
that the definition for the reference crop was a hypothetical reference crop with crop
height of 0.12 m, a fixed surface resistance of 70 s m − 1 and an albedo of 0.23
(Smith et al. 1992). The new equation is:
0.408 (Rn − G) + γ T +273

900
u 2 (es − ea )
E T0 = (12)
+ γ (1 + 0.34u 2 )
where E T0 = reference evapotranspiration rate (mm/day), T = mean air temper-

ature (°C), and u2 = wind speed (m s−1 ) at 2 m above the ground. This equation
can be applied using hourly data if the constant value “900” is divided by 24 for the
hours in a day and the Rn and G terms are expressed as MJ m−2 h−1 .
2.4 LandMOD ET Mapper
Bhattarai and Tao Liu (2019) have developed a new MATLAB-based ET mapping
toolbox named “LandMOD ET Mapper” for automated implementation of SEBAL
and METRIC to facilitate their widespread applications among new users with any
level of prior modeling experiences(Bhattarai and Liu 2019). The automation of
anchor pixels uses a modified version of Bhattarai et al. (2017) that eliminates the
need of having a reference weather station within the image.
The automation of hot and cold pixel starts with the selection of candidate pixels
based on a simple decision tree classifier and an exhaustive search used to select a
subset of pixels that could potentially be selected as hot and cold pixels. The final hot
and cold pixels are based on the ranking of each pixel based on its LST and NDVI
values. The pixel with the highest LST and lowest NDVI value is selected as the hot
pixel, while the pixel with the lowest LST and highest NDVI is taken as the cold
pixel. The automatically selected hot and cold pixels are used to internally calibrate
H (Sensible Heat Flux) and follow the stabilization correction procedures explained
in R. Allen et al. (2011). LE is then used to estimate evaporative fraction or reference
ET fraction (ETrF), which is assumed to be constant during the day to produce daily
ET maps, as in Allen et al.(2011) and (Bhattarai et al. 2017).
3 Results
3.1 Comparison of Daily Evapotranspiration

for Mymensingh Sadar Upazilla
In comparing the daily ET between SEBAL and METRIC algorithms (Table 2), it has
been found that METRIC ET are lower than SEBAL ET. The reason behind this can
be the internal calibration of two different methods. This study is limited to seven days
due to the unavailability of cloud-free Landsat images and measured data. We have
selected the dates when both cloud-free remote sensing images and measured ET
data were available. In comparison with SEBAL and METRIC ET against Penman–
Monteith method, we have noticed that the Penman–Monteith method shows higher
Table 2 Comparison of Daily ET estimated by SEBAL, METRIC, FAO Penman–Monteith

equation and Pan Evaporation methods for Mymensingh Sadar Upazilla
Serial Date of DOY SEBAL ET METRIC Pan Penman–Monteith
Nos 2007 (mm/day) ET Evaporation ET (mm/day)
(mm/day) (mm/day)
1 22 Jan 022 2.52 2.67 - 5.34
2 23 Feb 054 2.14 2.33 3.13 4.66
3 11 Mar 070 3.02 2.43 3.49 4.86
4 27 Mar 086 2.80 2.41 4.41 4.82
5 28 Apr 118 3.40 2.52 - 5.60
5 30 May 150 3.68 2.65 4.19 5.30
6 19 Sep 262 1.40 1.77 5.35 3.54
7 05 Oct 278 2.36 2.25 4.40 4.49
values than the SEBAL and METRIC models. This is expected because Penman–
Monteith provides potential ET but SEBAL and METRIC provide actual ET. On the
other hand, pan evaporation is comparable to Penman–Monteith ET. The actual ET
from the SEBAL and METRIC models have been found to be significantly correlated
with the Penman–Monteith (P-M) ET with a correlation coefficient of 0.86 and 0.95,
respectively, for the Mymensingh site (Fig. 4).
It has also been noticed that actual ET gradually increases in Boro season during
the growth stage. In fallow period, (between Boro and Aman cultivation) ET even-
tually decreases. Lower ET value in September indicates the initial stage of Aman
rice period. The spatial variation of daily ET from the models, across the study area,
is shown in Fig. 5. The values of ET vary from 0 to 7.35 mm/day. The lowest ET
have been observed from bare soil near the river bank and built up-areas (Fig. 5).
Mymensingh (SEBAL) Mymensingh (METRIC)

4 3
Actula ET from SEBAL
Actual ET from METRIC
3
(mm/day)
2
(mm/day)
2
1
1
R=0.86 R=0.95
0 0
3 4 5 6 3 4 5 6
Potenal ET from Penman-Monteith Potenal ET from Penman-Monteith
(mm/day) (mm/day)
Fig. 4 Comparison of the Penman–Monteith potential ET with the remote sensing-based estimates
of actual ET from the SEBAL and METRIC models for the Mymensingh site
Fig. 5 Comparison of daily ET between SEBAL and METRIC method for Mymensingh Sadar
Upazilla
3.2 Comparison of Daily Evapotranspiration for Bogra

Sadar Upazilla
The spatial and temporal variation of ET is shown in Fig. 6 for the Bogra station.
The values of ET vary from 0 to 5.65 mm/day near the BMD station. The lowest
ET have been observed from bare soil near the river bank and built up-areas. The
actual ET from the SEBAL and METRIC models have been found to be highly
correlated with the Penman–Monteith (P-M) ET. The correlation coefficient is 0.98
between the SEBAL and P-M method while it is 0.88 between the METRIC and
P-M method (Fig. 7). In comparison with daily ET between SEBAL and METRIC
Fig. 6 Comparison of daily ET between SEBAL & METRIC method for Bogra Sadar Upazilla
Bogra (SEBAL) Bogra (METRIC)

5 5
Actula ET from SEBAL
Actula ET from METRIC
4 4
(mm/day)
(mm/day)
3 3
2 2
1 1
R=0.98 R=0.88
0 0
3 4 5 6 7 3 4 5 6 7
Potenal ET from Penman-Monteith Potenal ET from Penman-Monteith
(mm/day) (mm/day)
Fig. 7 Comparison of the Penman–Monteith potential ET with the remote sensing-based estimates
of actual ET from the SEBAL and METRIC models for the Bogra Site
Table 3 Comparison of Daily ET estimated by SEBAL, METRIC, FAO Penman–Monteith

equation, and Pan Evaporation methods for Bogra Sadar Upazilla
Serial Date of DOY SEBAL ET METRIC Pan Penman–Monteith
Nos 2011 (mm/day) ET Evaporation ET (mm/day)
(mm/day) (mm/day)
1 24 Jan 24 1.88 2.85 - 3.21
2 09 Feb 40 3.21 3.41 - 5.08
3 25 Feb 56 3.61 3.62 - 5.42
4 13 Mar 72 4.72 5.64 - 6.32
5 16 May 136 4.06 4.35 5.1 6.20
6 08 Nov 312 1.89 2.26 3.3 3.56
algorithm, it has been found that METRIC ET is higher than the SEBAL ET, but
the difference is moderate. One reason behind this mismatch can be the internal
calibration of these two models. The differences in calculation of sensible heat in
these two models and the assumptions of SEBAL and METRIC in extrapolating
instantaneous ET to the daily ET may also contribute to the mismatch. Moreover
accuracy of ET estimation highly depends on the accuracy of weather data. Since no
actual weather data have been found available on an hourly scale, instead of actual
weather data, coarse resolution weather data from Merra 2 have been used. The
SEBAL and METRIC ET also differ from the pan evaporation. One reason behind
this difference can be the fact that no pan coefficient has been applied to get the
actual evaporation from measured evaporation. Besides these, the presence of
partial cloud shadow in some dates may also cause some errors in remote sensing-
based estimation (Table 3).
4 Conclusion
Evapotranspiration (ET) is a major component of the global water cycle. Measure-

ment of evapotranspiration is essential for regional water resources management, irri-
gation scheduling, hydrologic modeling as well as for global climate change studies.
The northwest region of Bangladesh depends highly on groundwater-based irrigation
during the dry season for cultivating Boro crop. An accurate estimation of ET can be
helpful in estimating irrigation demand precisely and can minimize the over extrac-
tion of groundwater in this region. In this study, the potential of SEBAL and METRIC
models in estimating ET have been assessed for two agricultural fields of Bangladesh.
A new MATLAB-based ET mapping toolbox named “LandMOD ET Mapper” has
been used for automated implementation of these models. SEBAL and METRIC
are more advantageous than conventional methods of evapotranspiration estimation
using crop coefficient curves or vegetation indices because SEBAL/METRIC models
do not require any specific crop or vegetation information, and they consider the effect
of water shortage, salinity or frost as well as evaporation from bare soil.
The spatial variation of ET estimated from the SEBAL and METRIC model have
been found to be comparable in this study. But the magnitudes are a little different.
This is probably due to the inherent differences in the algorithms of these two models.
The results show that the actual ET from SEBAL and METRIC are lower than the ET
from the FAO Penman–Monteith as the latter provides potential ET instead of actual
ET. This study used Landsat data. Future studies can be conducted by using MODIS
data instead of Landsat images. As MODIS acquire data on a daily basis, there is a
better chance of obtaining more cloud-free images. This study reveals a promising
potential of estimating ET from SEBAL and METRIC which can be very important
to better understand and address a variety of water resources related issues. There is a
need, however, to further test the model using additional datasets, and to examine the
performance of the model in other sites under different soil, climate, cropping, and
terrain conditions of Bangladesh. The results from this study are based on our initial
efforts to assess the operational characteristics and performance of the automated
SEBAL and METRIC model in the context of Bangladesh. Both models require reli-
able hourly weather data, including solar radiation, air temperature, relative humidity,
and wind speed, in order to compute reference evapotranspiration. Reference evap-
otranspiration is then used to compute and interpolate the evapotranspiration for the
days in between two satellite passes. The accuracy of this interpolation depends on
the quality of the dataset used to compute reference ET. Especially, in developing
countries like Bangladesh, hourly or even daily weather data are extremely limited.
Therefore, Merra-2 weather data have been used in this study as input in the SEBAL
and METRIC model. As measured weather data are not available on an hourly basis
in these sites, no validation of the dataset was possible. So, data from other global
gridded dataset can be used in the future as input in SEBAL and METRIC model
to check whether better results can be achieved. Both models have their operational
advantages and shortcomings depending on the conditions that are being applied. As
the SEBAL method requires lower number of ground measured weather data, it can
be said that it will be advantageous to use the SEBAL method over the METRIC
model for data-scarce locations like Bangladesh. Further research is needed to eval-
uate the practicality and value of both models in the context of Bangladesh to predict
surface ET throughout a growing season for different surfaces.
Acknowledgements The authors would like to express their gratitude to Nishan Bhattarai, Post-
doctoral Research Scientist, United States Department of Agriculture (USDA), for his immense
help with the LandMOD ET Mapper toolbox.
References
Acharjee, Tapos Kumar, Fulco Ludwig, Gerardo van Halsema, Petra Hellegers, and Iwan Supit.
2017. Future Changes in Water Requirements of Boro Rice in the Face of Climate Change in
North-West Bangladesh. Agricultural Water Management 194. Elsevier B.V.: 172–183. https://
doi.org/10.1016/j.agwat.2017.09.008.
Allen, R. G., L. S. Pereira, D. Raes, and M. Smith. 1998. FAO Irrigation and Drainage Paper No.
56—Crop Evapotranspiration, no. March.
Allen, Richard G., Masahiro Tasumi, Anthony Morse, Ricardo Trezza, James L. Wright, Wim
Bastiaanssen, William Kramber, Ignacio Lorite, and Clarence W. Robison. 2007. Satellite-Based
Energy Balance for Mapping Evapotranspiration with Internalized Calibration (METRIC)—
Applications. Journal of Irrigation and Drainage Engineering 133 (4): 395–406. https://doi.
org/10.1061/(asce)0733-9437(2007)133:4(395).
Allen, Richard, Ayse Irmak, Ricardo Trezza, Jan M.H.. Hendrickx, Wim Bastiaanssen, and Jeppe
Kjaersgaard. 2011. Satellite-Based ET Estimation in Agriculture Using SEBAL and METRIC.
Hydrological Processes 25 (26): 4011–4027. https://doi.org/10.1002/hyp.8408.
Anayah, F.M., and J.J. Kaluarachchi. 2014. Improving the Complementary Methods to Estimate
Evapotranspiration under Diverse Climatic and Physical Conditions. Hydrology and Earth System
Sciences 18 (6): 2049–2064. https://doi.org/10.5194/hess-18-2049-2014.
Anderson, M.C., W.P. Kustas, J.M. Norman, C.R. Hain, J.R. Mecikalski, L. Schultz, M.P. González-
Dugo, et al. 2011. Mapping Daily Evapotranspiration at Field to Continental Scales Using
Geostationary and Polar Orbiting Satellite Imagery. Hydrology and Earth System Sciences 15
(1): 223–239. https://doi.org/10.5194/hess-15-223-2011.
Bastiaanssen, W.G.M., M. Meneti, R.A. Feddes, and A.A.M. Holtslag. 1998. A Remote Sensing
Surface Energy Balance Algorithm for Land (SEBAL). Journal of Hydrology 212–213
(JANUARY): 198–212.
Bhattarai, Nishan, and Tao Liu. 2019. LandMOD ET Mapper: A New Matlab-Based Graphical
User Interface (GUI)for Automated Implementation of SEBAL and METRIC Models in Thermal
Imagery. Environmental Modelling and Software 118 (April). Elsevier: 76–82. https://doi.org/10.
1016/j.envsoft.2019.04.007.
Bhattarai, Nishan, Lindi J. Quackenbush, Jungho Im, and Stephen B. Shaw. 2017. A New Optimized
Algorithm for Automating Endmember Pixel Selection in the SEBAL and METRIC Models.
Remote Sensing of Environment 196. Elsevier Inc., 178–92. https://doi.org/10.1016/j.rse.2017.
05.009.
Bose, Indira, Faisal Hossain, Hisham Eldardiry, Shahryar Ahmad, Nishan K. Biswas, Ahmad
Zeeshan Bhatti, Hyongki Lee, Mazharul Aziz, and Md Shah Kamal Khan. 2021. Integrating
Gravimetry Data With Thermal Infra-Red Data From Satellites to Improve Efficiency of Opera-
tional Irrigation Advisory in South Asia.” Water Resources Research 57 (4). https://doi.org/10.
1029/2020WR028654.
Dey, Nepal C., Ratnajit Saha, Mahmood Parvez, Sujit K. Bala, A.K.M. Saiful Islam, Joyanta K.
Paul, and Mahabub Hossain. 2017. Sustainability of Groundwater Use for Irrigation of Dry-
Season Crops in Northwest Bangladesh. Groundwater for Sustainable Development 4: 66–77.
https://doi.org/10.1016/j.gsd.2017.02.001.
“Evapotranspiration in an Arid Environment: Quantifying the Moisture Input of Landscape Trees
and Turfgrass—ScienceDaily. 2021. Accessed August 19. https://www.sciencedaily.com/rel
eases/2020/12/201215164914.htm.
Hossen, Md Shahadat, Masayoshi Mano, Akira Miyata, Md Abdul Baten, and Tetsuya Hiyama.
2012. Surface Energy Partitioning and Evapotranspiration Over a Double-Cropping Paddy Field
in Bangladesh. Hydrological Processes 26 (9): 1311–1320. https://doi.org/10.1002/hyp.8232.
Kanoua, Wael, and Broder Merkel. 2015. Comparison Between the Two-Source Trapezoid Model
for Evapotranspiration (TTME) and the Surface Energy Balance Algorithm for Land (SEBAL)
in Titas Upazila in Bangladesh. FOG-Freiberg Online Geoscience 39: 65–86.
Karim, Fazlul, Mohammed Mainuddin, Masud Hasan, and Mac Kirby. 2020. Assessing the Potential
Impacts of Climate Changes on Rainfall and Evapotranspiration in the Northwest Region of
Bangladesh. Climate 8 (8). https://doi.org/10.3390/CLI8080094.
Kirby, J.M., M. Mainuddin, F. Mpelasoka, M.D. Ahmad, W. Palash, M.E. Quadir, S.M. Shah-
Newaz, and M.M. Hossain. 2016. The Impact of Climate Change on Regional Water Balances in
Bangladesh. Climatic Change 135 (3–4): 481–491. https://doi.org/10.1007/s10584-016-1597-1.
Li, Zhao Liang, Ronglin Tang, Zhengming Wan, Yuyun Bi, Chenghu Zhou, Bohui Tang, Guangjian
Yan, and Xiaoyu Zhang. 2009. A Review of Current Methodologies for Regional Evapotranspira-
tion Estimation from Remotely Sensed Data. Sensors 9 (5): 3801–3853. https://doi.org/10.3390/
s90503801.
Liou, Yuei An, and Sanjib Kumar Kar. 2014. Evapotranspiration Estimation with Remote Sensing
and Various Surface Energy Balance Algorithms-a Review. Energies 7 (5): 2821–2849. https://
doi.org/10.3390/en7052821.
Mojid, Mohammad A., Mohammed Mainuddin, Khandakar Faisal Ibn Murad, and John Mac Kirby.
2021. Water Usage Trends under Intensive Groundwater-Irrigated Agricultural Development in
a Changing Climate—Evidence from Bangladesh. Agricultural Water Management 251 (May).
Elsevier: 106873. https://doi.org/10.1016/J.AGWAT.2021.106873.
Monteith, John L. 1965. Evaporation and the Environment. 19th Symposia of the Society for
Experimental Biology 19: 205–234.
Murshed, Sonia Binte, and Jagath J. Kaluarachchi. 2018. Scarcity of Fresh Water Resources in the
Ganges Delta of Bangladesh. Water Security 4–5 (May). Elsevier: 8–18. https://doi.org/10.1016/
j.wasec.2018.11.002.
Murshed, Sonia Binte, Md Rezaur Rahman, and Jagath J. Kaluarachchi. 2019. Changes in Hydrology
of the Ganges Delta of Bangladesh and Corresponding Impacts on Water Resources. Journal of
the American Water Resources Association 55 (4): 800–823. https://doi.org/10.1111/1752-1688.
12775
Numata, Izaya, Kul Khand, Jeppe Kjaersgaard, Mark A. Cochrane, and Sonaira S. Silva. 2017.
Evaluation of Landsat-Based Metric Modeling to Provide High-Spatial Resolution Evapotran-
spiration Estimates for Amazonian Forests. Remote Sensing 9 (1). https://doi.org/10.3390/rs9
010046.
Penman, Howard Latimer. 1948. Natural Evaporation from Open Water, Bare Soil and Grass.
Proceedings of the Royal Society of London 193: 120–145. https://doi.org/10.1098/rspa.1948.
0037.
Reyes-González, Arturo, Jeppe Kjaersgaard, Todd Trooien, Christopher Hay, and Laurent
Ahiablame. 2017. Comparative Analysis of METRIC Model and Atmometer Methods for Esti-
mating Actual Evapotranspiration. International Journal of Agronomy 2017.https://doi.org/10.
1155/2017/3632501
Rojas, Jose P., and Ronald E. Sheffield. 2013. Evaluation of Daily Reference Evapotranspiration
Methods as Compared with the ASCE-EWRI Penman-Monteith Equation Using Limited Weather
Data in Northeast Louisiana. Journal of Irrigation and Drainage Engineering 139 (4): 285–292.
https://doi.org/10.1061/(asce)ir.1943-4774.0000523.
Ruhoff, A.L., A.R. Paz, L.E.O.C. Aragao, Q. Mu, Y. Malhi, W. Collischonn, H.R. Rocha, and S.W.
Running. 2013. Evaluation de l’algorithme MODIS d’estimation de l’évapotranspiration Globale
Utilisant Des Mesures de Covariance de La Turbulence et La Modélisation Hydrologique Dans
Le Bassin Du Rio Grande. Hydrological Sciences Journal 58 (8): 1658–1676. https://doi.org/10.
1080/02626667.2013.837578.
Shahid, Shamsuddin. 2011. Impact of Climate Change on Irrigation Water Demand of Dry Season
Boro Rice in Northwest Bangladesh. Climatic Change 105 (3–4): 433–453. https://doi.org/10.
1007/s10584-010-9895-5.
Smith, M., R.G. Allen, J.L. Monteith, L.S. Pereira, A. Perrier, and W.O. Pruitt. 1992. Report on the
Expert Consultation on Procedures for Revision of FAO Guidelines for Prediction of Crop Water
Requirements. Land and Water Development Division, United Nations Food and Agriculture
Service, Rome.
Stannard. 1993. Modified Priestley-Taylor Evapotranspiration Models. Water Resources 29 (5):
1379–1392.
Su, Z., and C. Jacobs (2001). Advanced Earth Observation: Land Surface Climate Final Report.
BCRS Report 2001: USP-2 Report 2001 01–02. Delft, Beleidscommissie Remote Sensing
(BCRS), 183.
Tasumi, M. 2003. Progress in Operational Estimation of Regional Evapotranspiration Using
Satellite Imagery. Ph.D. Dissertation. Moscow: University of Idaho.
Teixeira, A.H. de. C., Wim G.M. Bastiaanssen, Mobin-ud-Din Ahmad, and M.G. Bos. 2008.
Reviewing SEBAL Input Parameters for Assessing Evapotranspiration and Water Productivity for
the Low-Middle Sao Francisco River Basin, Brazil: Part A—Calibration and Validation. https://
cgspace.cgiar.org/handle/10568/40785.
Wang, Xiao Gang, Wen Wang, Dui Huang, Bin Yong, and Xi Chen. 2014. Modifying SEBAL Model
Based on the Trapezoidal Relationship Between Land Surface Temperature and Vegetation Index
for Actual Evapotranspiration Estimation. Remote Sensing 6 (7). MDPI AG: 5909–5937. https://
doi.org/10.3390/rs6075909.
Wilfried, Brutsaert. 1975. On a Derivable Formula for Long-Wave Radiation from Clear Skies FLD.
Water Resources Research 11 (5): 742–744.
Zhang, Ke, John S. Kimball, and Steven W. Running. 2016. A Review of Remote Sensing Based
Actual Evapotranspiration Estimation. Wiley Interdisciplinary Reviews: Water 3 (6): 834–853.
https://doi.org/10.1002/wat2.1168.
Zillman, J.W. 1972. A Study of Some Aspects of the Radiation and Heat Budgets of the Southern
Hemisphere Oceans/by J.W.|National Library of Australia. https://catalogue.nla.gov.au/Record/
2267389.

Gestion Del Agua

Uploaded by

Copyright:

Available Formats

Gestion Del Agua

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Gestion Del Agua

Uploaded by

Copyright:

Available Formats

G. M.

Water Management: A View

Ahmed Ishtiaque Amin Chowdhury

ISBN 978-3-030-95721-6 ISBN 978-3-030-95722-3 (eBook)

This book Water Management: A View from Multidisciplinary Perspectives presents

Dhaka, Bangladesh Prof. G. M. Tarekul Islam

Floods and Drainage

Hydrometeorological Hazards and Risk

Indigenous Knowledge and Practices of the Small Ethnic

Rivers, Coasts and Estuaries

Water Infrastructure and Development

Water and Livelihood Security

An Agent Based Model of Mangrove Social-Ecological System

Purnima Acharya and Ashutosh Shukla

Abstract The ever-growing incidences of urban flooding that were previously

Keywords Urban Flooding · LULCC · Climate Change · Storm water · Drainage

The annual flood brought by heavy monsoon precipitation results in enormous

P. Acharya (B) · A. Shukla

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 3

downstream. In addition, the changes in natural drainage and increased burden on

2.1 Study Area

Fig. 1 Study area

2.2 Reconstruction and Mapping of the Studied Flood Event

2.3 Reconstruction of Historical Events

Hanumante watershed is ungauged therefore longer-term hydrological data and

2.4 Meteorological Data Analysis

2.5 Map Analysis

2.6 Focus Group Discussion

2.7 Key Informants’ Interviews (KII)

Fourteen KIIs were conducted to collect information pertinent to plans, policies

2.8 Physical Observation of the Study Area

3.1 Historical Flood Events in Hanumante River

3.2 Analysis of Climatic Causes of the Flood

3.2.1 Rainfall Frequency Analysis

Table 1 Estimated 24-hours

Fig. 2 Likely 24 hours rainfall at different return periods in Bhaktapur

3.2.2 Analysis of Rainfall Trend

PRCPTOT: Trend of annual total precipitaƟon

RX1day: Trend of highest 1-day rainfall StaƟon

Rx5day: Trend of highest 5-day rainfall StaƟon

CWD: Trend of consecuƟve wet days

SDII: Simple daily intensity index

Fig. 3 Trend of rainfall analysis

3.2.3 People’s Perception on Climatic Trend

infrastructure development to be the additional cause. This attribution signifies that

3.3 Analysis of LULCC

3.3.1 LULCC of Kathmandu Valley

3.3.2 LULCC of Study Area

Fig. 4 Land Use Land Cover Change map of Kathmandu valley

Table 2 Land Use Changes in Kathmandu Valley (1988–2018)

Fig. 5 The site location for LULCC in MTM

Fig. 6 LULCC map around Hanumante River

Table 3 Land use changes in the study area over 1979–2018

3.3.3 People’s Observation on Urbanization vis-à-vis Trend of Flooding

3.4 Management of Flood

3.4.1 Adaptation and Mitigation Efforts Rolled at Household

3.4.2 Adaptation and Mitigation Rolled at Local Government Level

3.4.3 Adaptation and Mitigation Rolled at the Central Government

4.1 Sustainable Solutions to Mitigate Urban Flooding

Faruque Abdullah, A. K. M. Saiful Islam, Afsara Tasnia,